DEVELOPMENT OF AN AUTOLAND SYSTEM FOR GENERAL

AVIATION AIRCRAFT

Diana Siegel and R. John Hansman

This report is based on the S. M. Thesis of Diana Siegel submitted to the Department of
Aeronautics and Astronautics in partial fulfillment of the requirements for the degree of
Master of Science at the Massachusetts Institute of Technology.

The work presented in this report was also conducted

in collaboration with Avidyne, Inc

Report No. ICAT-2011-09

Sept 2011

MIT International Center for Air Transportation (ICAT)

Department of Aeronautics & Astronautics
Massachusetts Institute of Technology
Cambridge, MA 02139 USA
[Page Intentionally Left Blank]
 Development of an Autoland System for General Aviation
Aircraft
by
Diana Siegel

Submitted to the Department of Aeronautics and Astronautics

on September 28, 2011, in partial fulfillment of the
requirements for the degree of
Master of Science in Aeronautics and Astronautics

Abstract

Accidents due to engine failure, pilot disorientation or pilot incapacitation occur far more
frequently in general aviation than in commercial aviation, yet general aviation aircraft are
equipped with less safety-enhancing features than commercial aircraft. This thesis presents
the design of an emergency autoland system that includes automatic landing site selection,
guidance to the selected landing site and guidance along the final approach, in addition to
the automatic landing capability provided by conventional autoland systems. The proposed
system builds on the capability of a general aviation autopilot, flight management system and
GPS/WAAS augmented, integrated navigation system. The system provides this automatic
landing capability without the use of automatic throttle control and without the use of
a radar altimeter, which are essential to conventional autoland systems, but are typically
lacking on general aviation aircraft. The design addresses the challenge of no automatic
throttle control by utilizing only two simple power settings: cruise power and zero power.
The lack of radar altimeter is addressed by appropriate flare planning and placement of the
target touchdown point. The approach from the point of autoland initiation, to the approach
fix at the the landing site, is performed at cruise power, provided that power is available.
The final approach from the approach fix to touchdown, is performed at zero power. Control
of the touchdown point location during the final approach is achieved through adjustment
of the length of the trajectory, whenever the aircraft’s glide performance deviates from the
expected performance. The aircraft’s glide performance is measured online as the aircraft
tracks the planned trajectory. The performance of the final design is evaluated in simulation
in terms of touchdown point dispersion, sink rate and attitude on touchdown.

Thesis Supervisor: R. John Hansman

Title: Professor of Aeronautics and Astronautics

3
4
Acknowledgments

This work has been performed in collaboration with Avidyne, Inc. and was supported by
the Avidyne Fellowship.

5
6
Contents

1 Introduction 21

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.2 Motivation for Design of Autoland System . . . . . . . . . . . . . . . . . . . 22

1.3 Design Goal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

1.4 Design Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

1.5 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2 System Description 33

2.1 System Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.1.2 Engine Status Detection . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.1.3 Landing Site Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.1.4 Autoland Approach Procedure . . . . . . . . . . . . . . . . . . . . . . 39

2.1.5 Baseline Holding Pattern and Traffic Pattern . . . . . . . . . . . . . . 41

2.1.6 Trajectory Planning using Energy Management . . . . . . . . . . . . 44

2.1.7 Performance Prediction using Online Performance Estimation . . . . 48

7
 2.1.8 Aircraft Energy Monitoring and Energy Error Detection . . . . . . . 50

2.1.9 Trajectory Updating . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.1.10 Placement of the Target Touchdown Point . . . . . . . . . . . . . . . 63

2.1.11 Flare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

2.2 Autoland - FMS Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

2.2.1 Trajectory Specification . . . . . . . . . . . . . . . . . . . . . . . . . 71

2.2.2 Use of Local and Global Coordinate Systems . . . . . . . . . . . . . . 74

2.3 User Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

2.3.1 Autoland Initiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

2.3.2 Increase Throttle Command . . . . . . . . . . . . . . . . . . . . . . . 77

2.3.3 Status Display for Selected Landing Site and Planned Trajectory . . . 78

2.3.4 Prepare to Land Tutorial . . . . . . . . . . . . . . . . . . . . . . . . . 79

2.3.5 Request to Reduce Throttle at Initial Point . . . . . . . . . . . . . . 79

2.3.6 Request to Apply Brakes . . . . . . . . . . . . . . . . . . . . . . . . . 80

3 System Architecture 81

3.1 Static Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3.1.1 Autoland Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3.1.2 Autoland Components . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3.1.3 Autoland Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3.2 Autoland Modes and Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

3.2.1 Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

8
 3.2.2 Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

3.3 Autoland Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

3.3.1 External Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

3.3.2 Internal Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

3.3.3 Design Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

3.3.4 Complete System Diagram . . . . . . . . . . . . . . . . . . . . . . . . 103

4 Simulation Environment 105

4.1 Aircraft Model for Evaluation of Touchdown Point Dispersion . . . . . . . . 105

4.1.1 Point Mass Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4.1.2 Trajectory Tracking Controller . . . . . . . . . . . . . . . . . . . . . . 109

4.2 Aircraft Model for Evaluation of Flare Performance . . . . . . . . . . . . . . 111

4.2.1 High-fidelity Aircraft Model . . . . . . . . . . . . . . . . . . . . . . . 111

4.2.2 Flare Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

4.3 Wind Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4.4 Wind Estimator Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

4.5 Performance Estimator Model . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4.6 Energy Manager Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

4.7 Trajectory Updator Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

5 Results 117

5.1 Evaluation of Effect of Wind on Touchdown Point Dispersion . . . . . . . . . 118

9
 5.1.1 Test Setup For Evaluation of Effect of Wind on Touchdown Point
Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.1.2 Illustration of Test Case 1 for Evaluation of Effect of Wind on Touch-

down Point Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . 120

5.1.3 Summary of Effect of Wind on Touchdown Point Dispersion . . . . . 125

5.1.4 Discussion of Effect of Wind on Touchdown Point Dispersion . . . . . 126

5.2 Evaluation of Flare Performance . . . . . . . . . . . . . . . . . . . . . . . . . 127

5.2.1 Test Setup for Evaluation of Flare Performance . . . . . . . . . . . . 127

5.2.2 Illustration of Test Case 1 for Evaluation of Flare Performance . . . . 128

5.2.3 Summary of Flare Performance Test Results . . . . . . . . . . . . . . 131

5.2.4 Discussion of Flare Performance Test Results . . . . . . . . . . . . . 132

6 Conclusion and Future Work 135

6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

A Module Descriptions 141

A.1 Modules of Layer Destination Planning . . . . . . . . . . . . . . . . . . . . . 141

A.1.1 Engine Status Detection . . . . . . . . . . . . . . . . . . . . . . . . . 141

A.1.2 Range calculator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

A.1.3 Landing Site Prioritizer . . . . . . . . . . . . . . . . . . . . . . . . . 144

A.1.4 Initial Point Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . 149

A.1.5 Touchdown Energy Calculator . . . . . . . . . . . . . . . . . . . . . . 150

10
A.2 Modules of Layer Trajectory Planning . . . . . . . . . . . . . . . . . . . . . 152

A.2.1 Energy Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

A.2.2 L/D 0 estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

A.2.3 L/D expected Generator . . . . . . . . . . . . . . . . . . . . . . . . . 157

A.2.4 Path planner to Initial Point . . . . . . . . . . . . . . . . . . . . . . . 159

A.2.5 Flight Plan Feasibility Monitor . . . . . . . . . . . . . . . . . . . . . 161

A.2.6 Excess Energy Calculator . . . . . . . . . . . . . . . . . . . . . . . . 163

A.2.7 Holding Pattern Generator/Updator . . . . . . . . . . . . . . . . . . 165

A.2.8 Traffic Pattern Trajectory Generator . . . . . . . . . . . . . . . . . . 167

A.2.9 Energy Curve Manager . . . . . . . . . . . . . . . . . . . . . . . . . . 171

A.2.10 Traffic Pattern Energy Error Detector . . . . . . . . . . . . . . . . . 174

A.2.11 Traffic Pattern Trajectory Updator . . . . . . . . . . . . . . . . . . . 176

A.3 Modules of Layer Trajectory Following . . . . . . . . . . . . . . . . . . . . . 187

A.3.1 Trajectory to Initial Point Follower . . . . . . . . . . . . . . . . . . . 187

A.3.2 Holding Pattern Follower . . . . . . . . . . . . . . . . . . . . . . . . . 188

A.3.3 Traffic Pattern Follower . . . . . . . . . . . . . . . . . . . . . . . . . 191

A.3.4 Flare Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

A.4 Selected Library Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

A.4.1 convertToGlobalCoords . . . . . . . . . . . . . . . . . . . . . . . . . . 195

A.4.2 getTurnEntryLine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

A.4.3 lineCrossed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

11
A.4.4 calcDistToGo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

A.4.5 calcInterceptPoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

A.4.6 calcTurnDist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

12
List of Figures

1-1 Autoland integration with its environment . . . . . . . . . . . . . . . . . . . 26

1-2 Airspeed and flight path angles for various throttle settings . . . . . . . . . . 28

2-1 Overview of autoland concept . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2-2 Pseudocode for automatic engine status detection . . . . . . . . . . . . . . . 36

2-3 Landing site selection process . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2-4 Aircraft trajectory during autoland for case engine operational and no obstacles 40

2-5 Sequence of functions performed during approach procedure in case engine

operational . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2-6 Length and shape of the baseline trajectory . . . . . . . . . . . . . . . . . . 43

2-7 Naming convention for holding pattern legs and traffic pattern legs . . . . . 46

2-8 Energy dissipation with ground distance travelled . . . . . . . . . . . . . . . 47

2-9 Determination of observed glide ratio while in flight . . . . . . . . . . . . . . 49

2-10 Calculation of the expected glide ratio, L/De xp, from L/D0 . . . . . . . . . 50

2-11 Concept of excess energy correction through trajectory elongation . . . . . . 51

2-12 Update of energy vs. range-to-go reference curve upon receipt of new aircraft
glide performance and wind estimate . . . . . . . . . . . . . . . . . . . . . . 52

13
2-13 Direction and sizing of holding pattern . . . . . . . . . . . . . . . . . . . . . 54

2-14 Trajectory updating method on the downwind leg . . . . . . . . . . . . . . . 56

2-15 Trajectory updating method on the base leg . . . . . . . . . . . . . . . . . . 58

2-16 Example of trajectory updating on the base leg after first update . . . . . . 59

2-17 Concept of a 3D lookup table for determination of d on downwind leg . . . . 61

2-18 Percent change in length of trajectory and amount of final leg length required
for range of angles turned away from trajectory . . . . . . . . . . . . . . . . 62

2-19 Example of change in wind speed, S and direction, dir with altitude in the
atmospheric boundary layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

2-20 Linear approximation of change in wind speed in the boundary layer . . . . . 65

2-21 L/D error resulting from assuming linear reduction in wind speed of Vw,0 when
wind reduces only by ∆Vw . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

2-22 Overshoot or undershoot from error in estimating expected L/D . . . . . . . 68

2-23 Placement of target touchdown point for consideration of L/D estimation error
and GPS error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

2-24 Flare maneuver under uncertainty in vertical position . . . . . . . . . . . . . 71

2-25 Specification of the autoland trajectory as set of straight line segments and
hold loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

2-26 Local North-East coordinate system before and after rotation by amount equal
to runway heading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

2-27 Conceptual visual cue for operator to increase throttle . . . . . . . . . . . . 78

2-28 Conceptional display of target destination and planned trajectory on moving

map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

14
2-29 Conceptual visual cue for operator to reduce throttle to zero . . . . . . . . . 79

3-1 Autoland layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3-2 Autoland system architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3-3 Autoland system architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3-4 Autoland modes as determined by engine status . . . . . . . . . . . . . . . . 90

3-5 Autoland phases determining the lateral trajectory followed by the aircraft . 93

3-6 Autoland external interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

3-7 Autoland component internal interfaces active during phase ”Follow Traffic
Pattern” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

3-8 Autoland component internal interfaces . . . . . . . . . . . . . . . . . . . . . 99

3-9 Full system diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

4-1 Wind profile for linear reduction in wind speed in boundary layer . . . . . . 114

4-2 Wind profile for near-constant wind speed in boundary layer . . . . . . . . . 115

4-3 Wind profile for inversion of wind direction in boundary layer . . . . . . . . 116

5-1 Left traffic pattern trajectory used to test the performance of the autoland
software prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

5-2 Traffic pattern after update from 100 ft altitude error . . . . . . . . . . . . . 120

5-3 Energy vs. range-to-go curve before first update . . . . . . . . . . . . . . . . 121

5-4 Energy vs. range-to-go curve before first update, enlarged . . . . . . . . . . 121

5-5 Energy vs. range-to-go curve after first update . . . . . . . . . . . . . . . . . 122

5-6 Energy vs. range-to-go curve after first update, enlarged . . . . . . . . . . . 122

15
5-7 Traffic pattern after second update due to reduction in L/D0 . . . . . . . . . 122

5-8 Energy vs. range-to-go curve before second update . . . . . . . . . . . . . . 123

5-9 Energy vs. range-to-go curve before second update, enlarged . . . . . . . . . 123

5-10 Energy vs. range-to-go curve after second update . . . . . . . . . . . . . . . 123

5-11 Energy vs. range-to-go curve after second update, enlarged . . . . . . . . . . 123

5-12 Trajectory before and after each update and actual trajectory flown . . . . . 124

5-13 Test setup for performance evaluation of flare controller . . . . . . . . . . . . 128

5-14 Commanded sink rate and actual aircraft sink rate during flare maneuver . . 129

5-15 Altitude vs. horizontal distance travelled during flare maneuver . . . . . . . 130

5-16 Change in aircraft position and velocity during flare maneuver . . . . . . . . 130

5-17 Evolution of angle of attack, flight path angle and pitch during flare maneuver 131

6-1 Autoland integration with its environment . . . . . . . . . . . . . . . . . . . 138

A-1 Runway search in engine operational case . . . . . . . . . . . . . . . . . . . . 146

A-2 Runway search algorithm in engine operational case . . . . . . . . . . . . . . 147

A-3 Landing site selection process for case Engine Out . . . . . . . . . . . . . . . 148

A-4 Pattern direction from cross product of V1 and V2 . . . . . . . . . . . . . . 150

A-5 Inferring tailwind component from ground speed and wind speed . . . . . . . 155

A-6 Effect of wind on horizontal distance travelled per unit of time . . . . . . . . 155

A-7 Nodes and edges connecting the nodes on a visibility graph . . . . . . . . . . 161

A-8 Distance, d, to move base leg fix to correct energy error ∆E/W . . . . . . . 169

16
A-9 Baseline and updated trajectory for max. and min. energy error expected
(rotated by −90◦ from local coordinate system) . . . . . . . . . . . . . . . . 169

A-10 Local energy vs. range curve function representation . . . . . . . . . . . . . 173

A-11 Energy error tolerance bounds . . . . . . . . . . . . . . . . . . . . . . . . . 174

A-12 d vs. E/W plot for various L with αp = const = 45◦ . . . . . . . . . . . . . . 178

A-13 d vs. E/W plot for various αp with L=const = 5,000 ft . . . . . . . . . . . . 179

A-14 Geometry used to determine difference in path length between the current
and updated trajectory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

A-15 d vs. E/W plot for various L with dtotal =500 ft and αp = const = 30◦ . . . . 183

A-16 d vs. E/W plot for various αp with dtotal = 500f t and L = 3, 000f t . . . . . . 183

A-17 d vs. E/W plot for various dtotal with αp = 45◦ and L = 3, 000f t . . . . . . . 184

A-18 d vs. E/W plot for extreme ∆E/W to illustrate corner cases . . . . . . . . . 185

A-19 Geometry used to determine location of IP and OP . . . . . . . . . . . . . . 190

A-20 Calculation of turn entry line . . . . . . . . . . . . . . . . . . . . . . . . . . 197

A-21 Geometry used to determine whether turn entry line has been crossed at
location, P . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

A-22 Calculation of Range-To-Go until start of the next segment . . . . . . . . . . 202

A-23 Calculation of distance from line . . . . . . . . . . . . . . . . . . . . . . . . . 203

A-24 Geometry used to determine the lead distance to the waypoint at which a
turn should be initiated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

17
18
List of Tables

1.1 GA accidents in 2007 and 2008 . . . . . . . . . . . . . . . . . . . . . . . . . 23

1.2 Requirements on autoland system . . . . . . . . . . . . . . . . . . . . . . . . 27

3.1 Autoland external interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

3.2 Autoland internal interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

3.3 Autoland design constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

5.1 Summary of test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

5.2 Summary of flare performance test results . . . . . . . . . . . . . . . . . . . 132

19
20
Chapter 1

Introduction

1.1 Background

Automatic landing systems have existed for about half a century, with British Airways having
made the first automatic landing during a commercial flight in June 1965 using an auto-
matic flare controller in the pitch axis [1]. A landing with full control in all three axes was
achieved shortly after in 1966. Landing aids that guide the aircraft down to flare height have
been developed since 1920 [16]. On September 19, 1930, H. Diamond and F.W. Dunmore
presented the design and hardware for a landing aid based on radio navigation similar to
today’s Instrument Landing System (ILS) [9]. Today’s commercial autoland systems utilize
the ILS’ vertical and lateral guidance to descend towards the runway at a glide path of −3◦ ,
while keeping the aircraft’s track aligned with the runway centerline. At around 30 ft above
the ground, the aircraft enters the flare mode, during which the aircraft pitches up to reduce
the vertical speed from its current descent rate to 1-3 ft/s by touchdown [18].
Automatic landing equipment is certified to either category CAT I, CAT II or CAT III,
which determines the minimum altitude above the touchdown zone that the automatic land-
ing system may be used. This minimum altitude is defined as the decision height. At the
decision height, the pilot must have the runway in sight and be able to see as far down the

21
runway as prescribed by the minimum runway visual range. If, at the decision height, the
runway is not in sight, or runway visual range is smaller than the minimum range specified,
the pilot is required to execute a missed approach and go-around to re-attempt the landing
or divert to another airport. In order to be able to perform an approach in a specific cat-
egory, both the on-board and the ground-based equipment need to be certified to at least
this specific category.
General aviation aircraft level equipment is typically certified to CAT I only, which spec-
ifies a decision height not lower than 200 ft and a runway visual range not less than 1,800
ft. Therefore, general aviation autopilot systems must be decoupled at 200 ft and the land-
ing performed manually. Commercial aircraft are typically certified to CAT III, which is
sub-divided into CAT IIIa, CAT IIIb and CAT IIIc. CAT IIIc specifies zero decision height
and zero runway visual range and is to the author’s knowledge at this time not certified
for any ground-based equipment at any airport worldwide. CAT IIIb authorizes landings
with as low as zero decision height and runway visual range as low as 150 ft. The CAT
IIIb ILS approach to runway 25L at LAX, for example, is certified for zero decision height
and runway visual range of 600 ft. Certification to CAT III requires redundant auto-flight
systems including redundant ways of actuating all flight controls and redundant autopilots.
Furthermore redundant radar altimeters are required to accurately measure the distance over
ground. This type of equipment is commonly found on commercial aircraft but not typically
on general aviation aircraft since the hardware cost is prohibitive for most private operators.

1.2 Motivation for Design of Autoland System

Accidents in commercial aviation have continuously decreased since the early days of avia-
tion with typical fatal accident rates now being less than 0.02 per 100,000 flight hours with
a total of 16 million flight hours [22]. The fatal accident rate in general aviation, however, is
more than 50 times higher than that, with a rate of 1.3 per 100,000 flight hours for a total
of 30 million flight hours [6]. In part, this is due to the fact that general aviation pilots fly

22
 2007 2008
Total (fatal) Total (fatal)
Mechanical failure (with power loss) 132 (14) 104 (9)
Unknown loss of power 118 (12) 102 (14)
Fuel mismanagement 90 (9) 73 (9)
Pilot incapacitation 6 (4) 3 (3)
Total loss of power +
pilot incapacitation accidents 346 (39) 283 (35)
Total number of GA accidents 1385 (252) 1254 (236)
Loss of power +
pilot incapacitation % of all GA accidents 25.0% (15.5%) 22.6% (14.8 %)

Table 1.1: GA accidents in 2007 and 2008

shorter routes and hence have more takeoffs and landings, the most accident-prone part of
the flight, compared to commercial operators. The difference in pilot training and aircraft
equipage, however, also contributes to the high accident rates in general aviation compared
to commercial aviation. Commercial aircraft are equipped with many safety enhancing fea-
tures, such as traffic advisory systems, ground proximity warning systems, etc. that general
aviation aircraft lack. Given the accident statistics, however, there is an equal need for ad-
ditional safety enhancing features on general aviation aircraft.
Accidents in general aviation occur due to mechanical failures, adverse weather and pi-
lot error to name a few. This thesis addresses two specific causes for accidents in general
aviation, which taken together produce a large proportion of the fatal accidents: pilot in-
capacitation and loss of power. Though pilot incapacitation contributes few numbers of
accidents, the chance that this type of accident ends fatal, is much higher than for any
other type of accident. Pilot incapacitation may occur due to hypoxia from excess altitude,
or any other medical condition. Loss of power is either due to mechanical failure or fuel
mismanagement, which occurs surprisingly often, despite being entirely preventable. Table
1.1 summarizes accident statistics for general aviation aircraft in 2007 and 2008. Accidents
due to loss of power and pilot incapacitation amounted to 25.0% and 22.6% of all general
aviation accidents in 2007 and 2008 respectively [3, 2].
In case of pilot incapacitation, the pilot is either completely or at least partially inca-

23
pable to maintain control of the aircraft. In case of loss of power, the achievable flight range
is heavily reduced, which makes it difficult for the pilot to judge which landing sites are
reachable. Hence, providing automatic control of the aircraft’s trajectory and selecting and
guiding the aircraft to a reachable landing site, are two main objectives that an autoland
system aimed at the scenarios of pilot incapacitation and loss of power will need to address.
There would be no obstacles to implementing this kind of system, if autoland had ac-
cess to a representation of the aircraft’s environment and location of landing sites, and had
automatic control over all flight control surfaces and the throttle. The autopilot and flight
management systems as commonly found on commercial aircraft, provide precisely this func-
tionality. A decade ago, these systems were rarely found on general aviation aircraft, but
they are starting to be integrated more and more on this class of aircraft.
For example, the Cirrus SR-22 incorporates a complete flight automation solution with
autopilot and flight management system. This system allows full control of the aircraft’s
motion in the lateral plane, but only limited control in the vertical plane, since this class
of aircraft typically lacks automatic throttle control. Furthermore, its integrated naviga-
tion systems (INS) features WAAS (wide area augmentation system) enabled GPS receivers,
which provide a much better position accuracy than previous GPS receivers. Without Se-
lective Availability, which was turned off in 2000, previous GPS receivers achieved a typical
accuracy of 15 m laterally and 20 m vertically. This level of accuracy was clearly insufficient
for the landing task. The WAAS specification requires that a position accuracy of better
than 7.6 m be achieved both laterally and vertically at least 95% of the time. However,
measurements in the continental US have shown that GPS/WAAS receivers achieve a typ-
ical accuracy of < 2 m laterally and vertically, which provides the potential to perform a
precision approach based on the WAAS position information only [4]. This improvement in
the accuracy of position sensing is important, since general aviation aircraft typically do not
possess a radar altimeter, which provides commercial aircraft with centimeter level vertical
position accuracy during the final stage of the landing.

24
1.3 Design Goal

The aim of this thesis is to design an autoland system to be used in case of emergency.
Specifically, the design addresses the following scenarios:

• Partial or complete pilot incapacitation

• Loss of power

Since loss of power can occur at any point during the flight and the pilot can become
incapacitated at any point in time, the autoland system to be developed needs to have a
larger scope than the typical autoland systems on commercial aircraft, which only take over
once the aircraft is aligned with the runway centerline on final approach. For the case of
partial or complete pilot incapacitation, the pilot needs to be able to rely on autoland to
automatically select a suitable landing site, guide the aircraft to the landing site and perform
the approach and landing. For complete pilot incapacitation, autoland may not require any
input from the pilot. However, there is the possibility of a passenger being present, which
autoland aims to draw on, in order to provide control inputs to enhance autoland performance
in the event of complete pilot incapacitation. This passenger is assumed to be untrained in
piloting the aircraft, so that any interaction with the passenger should be kept as simple as
possible. The same applies to any interaction with a partially incapacitated pilot.
For the case of loss of power, the pilot can assumed to be alert. However, in this situation,
it is difficult for the pilot to judge which landing sites are reachable, how to best reach the
landing site given en-route terrain and how to match the length of the final approach to the
glide performance of the aircraft. Hence, automatic landing site search and guidance to the
landing site is required in this scenario, in addition to guidance along the final approach.
This capability will need to be achieved with only the typical hardware equipage found
on a general aviation aircraft. Adding hardware is considered to not be an option, since the
cost would inhibit adoption of the system. The model aircraft used to represent a ”typical”
general aviation aircraft is a Cirrus SR-22 equipped with an Avidyne DFC-90 autopilot,

25
GPS/WAAS augmented integrated navigation system (INS) and a flight management system
(FMS). It is acknowledged that many general aviation aircraft do not possess this level of
automation. However, this design will be targeted at aircraft that are already equipped with
an autopilot, FMS and INS.
Figure 1-1 shows the environment, in which autoland is expected to operate. For landing

Operator

autoland
status,
commands
aileron

elevator
Autoland
throttle

mode
rudder
navigation lateral
select aircraft brakes
data trajectory state

flaps
Flight Management System

roll steering commands

Autopilot

aileron
elevator

Aircraft

Figure 1-1: Autoland integration with its environment

site selection and trajectory planning, autoland can draw aircraft state information, such as
aircraft location, velocity and angular rates from the autopilot, and navigation data from
the FMS. After having constructed a path to the selected landing site, autoland can pass the
desired trajectory to the FMS for tracking in the lateral plane. Based on the commanded
trajectory and the aircraft’s current position, the FMS generates roll steering commands,
that the autopilot subsequently passes on to the aileron servo. The autopilot itself controls

26
 Functional requirements
FUNC 1 Select a suitable landing site
FUNC 2 Compute an obstacle-free trajectory to the
selected landing site
FUNC 3 Guide the aircraft to the selected landing site
FUNC 4 Compute a feasible approach pattern to
the touchdown zone
FUNC 5 Guide aircraft along the approach pattern
FUNC 6 Perform automatic touchdown
FUNC 7 Indicate autoland status to pilot and passenger
Non-functional requirements
NFUNC 1 Implement functionality as software upgrade only
(no additional hardware to be included)
NFUNC 2 Commands given to the operator shall be simple
enough to require no prior pilot training
NFUNC 3 Design system interface to the FMS and the autopilot
so that little to no modification of the FMS/autopilot
functionality is required

Table 1.2: Requirements on autoland system

the elevator based on the selected vertical mode. Aileron and elevator may be actuated
manually or automatically, as illustrated by the connector from the operator to the aircraft
and from the autopilot to the aircraft. The throttle, rudder, brakes and flaps, however,
may only be actuated manually. This imposes natural limitations on the performance of
the autoland system for general aviation aircraft when compared to the performance of
the autoland system of a commerical aircraft, which also possesses automatic control over
rudder, throttle and brakes. In addition to that, autoland will need to perform the touchdown
without a radar altimeter. A summary of the requirements on the autoland system to be
developed is given in table 1.3.

1.4 Design Constraints

The lack of auto-throttle has the strongest effect on autoland performance, since it limits
the achievable flight path angles. Figure 1-2 shows the limitations imposed by the lack of
throttle control, by illustrating the range of airspeeds and flight path angles achievable for a

27
given throttle setting. The throttle setting ranges from 0 to 1 representing a change in engine
power output from 0% to 100% ignoring losses and non-linear behavior. If the throttle is
stuck in one position, the aircraft’s motion in the vertical plane can only be effected by the
elevator. The elevator can be used to increase or decrease the flight path angle and increase
and decrease the airspeed, but cannot control both the flight path angle and the airspeed
at the same time. As a result, the elevator can increase the flight path angle only until the
airspeed reaches the stall speed, and decrease the flight path angle only until the airspeed
exceeds the aircraft’s maximum design speed. Hence, the lack of auto-throttle limits the
range of flight path angles achievable for a given throttle setting.
If the pilot became incapacitated during cruise flight, i.e. 0◦ flight path angle, the throttle

cruise airspeed

approach
throttle setting

approach speed
cruise power
setting

Figure 1-2: Airspeed and flight path angles for various throttle settings

would typically be fixed between 50% and 70%. The airspeed would hence measure between
140 - 160 kt. The standard ILS glideslope guides the aircraft along a flight path angle of
−3◦ . At a throttle setting of 60%, this flight path angle would result in an airspeed of
170 kt, which is excessive given the recommended approach speed of 90 kt for this aircraft.

28
Approaching at this speed would result in the nose wheel hitting the ground first, if an
attempt to land were made, or the aircraft floating off the runway and regain altitude, if the
nose was attempted to be pulled up before touchdown. In order to avoid stall or excessive
airspeed during final approach, the throttle setting should be kept within 6% to 8%. It
should be recognized, that manually setting the throttle to this very narrow region is a non-
trivial task, especially when the exact numbers will need to be adjusted for wind and aircraft
performance from one landing to another. Assuming the pilot is incapacitated, a passenger
would most likely not be able to establish the correct throttle setting in time. Therefore, the
autoland system to be developed, is forced to consider alternatives to the standard approach
on a −3◦ glideslope. During the loss of power scenario, the throttle setting can be taken as
effectively zero. Inspection of Figure 1-1 shows that the flight path angle at the recommended
approach speed of 90 kt is −6◦ .
The lack of radar altimeter is considered to have the second strongest effect on autoland
performance. Not knowing the exact time and location, at which the aircraft will touch
down, leads to touchdown point dispersion and the requirement on autoland to plan the
flare so that the aircraft an earlier, or later than expected touchdown can be tolerated.
The lack of rudder affects autoland’s ability to ”de-crab” the aircraft during a crosswind
landing before touchdown, i.e. aligning the nose of the aircraft with the runway centerline.
Touching down sideways with the nose pointing off the runway centerline, would exert an
undesirable load on the landing gear and potentially cause the aircraft to run off the side of
the runway. However, this is deemed to be tolerable, given that the system is aimed at an
emergency situation.
Finally, if neither the pilot nor the passenger are able to operate the brakes, the lack of
automatic braking would cause an elongation in rollout distance of at minimum a factor of
five, which would cause the aircraft to run off the end of the runway. This is considered
to be a dangerous scenario, since obstacles are likely to be present beyond the end of the
runway. Hence the pilot or passenger would need to be requested to apply brake pressure,
if at all possible. If no manual input can be provided, autoland would need to exercise any

29
means available to decrease the stopping distance. One possibility is to deliberately run the
aircraft off the side of the runway to utilize the higher rolling friction provided by grassy or
dirt areas.
It should be noted that, if the aircraft autoland is designed for, uses a retractable gear,
the pilot would need to be requested to lower the gear before the approach is started. Since
the Cirrus SR-22 is fitted with non-retractable landing gear, this step is not necessary.

1.5 Literature Review

Emergency landing aids have previously been investigated for application to commercial
aircraft. Meuleau, Plaunt and Smith proposed an emergency landing planner for damaged
aircraft, which determines potential landing sites and feasible trajectories to the landing site
and organizes each solution by increasing amount of risk. The decision to which airport
to fly along which route is left to the pilot [15]. Furthermore, Atkins, Portillo and Strube
presented an emergency flight planner specifically applied to total loss of thrust for general
aviation aircraft. This planner calculates the aircraft’s reachable footprint and uses prioriti-
zation based on information from the FMS database to automatically find the most suitable
landing site. Subsequently, the planner generates a trajectory to the touchdown zone by
using estimates on the minimum and maximum achievable vertical flight path angle. The
planner is adaptive in the sense that it updates the projected trajectory based on changes
in aircraft performance [7]. The space shuttle’s mission abort flight manager (SAFM) uses a
similar approach. Based on the current state vector of the shuttle, the SAFM determines a
feasible landing site and calculates a trajectory to this landing site using aircraft performance
boundaries. These boundaries are determined from knowledge of the maximum and mini-
mum drag profiles achievable by the control system [14]. The space shuttle’s terminal area
energy manager (TAEM) synthesizes a landing trajectory from pre-defined maneuvers and
controls energy vs. range-to-go until touchdown by modulating dynamic pressure and speed
brake. If these modulation tequniques are insufficient, the TAEM elongates or shortens the

30
trajectory to match the observed shuttle energy [10].
Controllers that guide the aircraft along a suitable vertical profile once the aircraft has
been aligned with the runway centerline have been designed for commercial aircraft, small
and large-scale Unmanned Aerial Vehicles (UAV) and also for general aviation sized aircraft.
In all cases, automatic control over thrust, however, has been assumed. Wagner and Valasek,
for example, investigated the performance of a QFT controller in comparison to the standard
proportional-integral controller for automatic landing of a medium-sized UAVs and general
aviation aircraft [21]. Shue and Agarwal designed an H2/H-Infinity controller for automatic
landing of commercial aircraft [19].
In 1994, the FAA and UPS successfully tested automatic landings using only differential
GPS (DGPS) for positioning information with a Boeing 757 [13]. DGPS uses local ground
stations installed at the airport to transmit corrections for the raw GPS signal received from
the satellite. Hence, given accurate position information and automatic throttle control,
automatic landing controllers have been proven successful for any size of aircraft from small
UAVs to large transport aircraft. The specific contribution of this thesis, is to investigate the
feasibility of an automatic landing system for general aviation aircraft without automatic
throttle control and positioning information only to the accuracy of WAAS augmented GPS.

31
32
Chapter 2

System Description

2.1 System Concept

2.1.1 Overview

The autoland system proposed encompasses the tasks from autoland initiation, automatic
scenario detection, landing site selection and trajectory planning, to final approach and
landing. The autoland system resolves the challenge of no automatic throttle control by
using only two simple, distinct power settings: cruise power and zero power. Cruise power is
used to fly from the point of autoland initiation to an initial point located at a fixed distance
and direction from the runway threshold of the selected landing site. Zero power is then used
to fly the final approach from the initial point to the touchdown point. This choice is due to
the fact that at zero power, it is always possible to perform a landing without further input
from the pilot, assuming that initial conditions and wind speed remain within design range.
If the autoland system attempted to perform an approach on the standard −3◦ glideslope,
the system would need to rely on the pilot to reduce the throttle to zero at the time of flare
in a timely manner, or else the aircraft would be trapped in a situation, where it is unable to
climb in order to abort the landing, and unable to land due to excess power. Excess power

33
on landing would either result in the nose-wheel touching down first or the aircraft floating
off the runway. Furthermore, requesting the operator to reduce the throttle all the way to
zero is a simple command to follow. It is reasonable to assume that a partially incapacitated
pilot or untrained passenger will be able to follow a command requesting to pull the throttle
back when given visual and audible cues. Finally, this approach has the advantage that the
”Incapacitated Pilot” and ”Loss of Power” scenarios can be treated equally from the start
of the gliding approach at the initial point. The two cases need to be treated differently
only for landing site selection and guidance to the initial point. Figure 2-1 summarizes the
overall design concept.
After initiation, autoland automatically determines whether the engine is operational or
has failed. Based on the engine status, autoland selects a landing site within the aircraft’s
range and guides the aircraft to the initial point. At the initial point, autoland requests the
pilot to reduce the throttle to zero, if the power is not zero already, so that the final approach
can be performed at power off. With this choice, the timing of the throttle reduction is not
critical, which reduces reliance on the human operator. As long as the power has not been
reduced to zero, autoland commands the aircraft to loiter in a holding pattern at the initial
point. With this approach, the system is also capable of fully autonomous operation, given
that the initial throttle setting is sufficient for straight and level flight. In the event that the
pilot is fully incapacitated and cannot actuate the throttle, the aircraft loiters until fuel is
depleted. Once the power has reduced to zero due to fuel starvation, the approach is started.
With this concept, the autoland system is capable of autonomously guiding the aircraft
from its current location to the selected landing site given that the throttle setting is greater
than the setting required to maintain straight and level flight, or smaller than the maximum
power at which a successful power on approach can be performed. If the power setting is in
between these thresholds, the aircraft will not be able to hold altitude, while not being able
to land in a nose-up attitude. In this situation, autoland will audibly and visually request
the operator to increase throttle, while aiming for an unpopulated area at wings level. As
soon as autoland has detected an increase in power, re-planning will be performed in order

34
 Autoland initiation
by button press

Automatic engine
status detection

Engine operational Engine failed

Large number
Landing site of possible
Limited number of
selection landing sites ? possible landing sites
based on range

Reduce power
to zero

Final approach
with power-off

Runway Runway

Figure 2-1: Overview of autoland concept

to select a more favorable emergency landing site.

2.1.2 Engine Status Detection

Engine status detection is performed automatically using the %power indicator available
on the Cirrus SR-22. The engine status will either be classified as ”Engine Operational”,
”Engine Out” or ”Indeterminate”. The engine status is deduced to be ”Engine Operational”
if the engine output power is high enough so that straight and level flight can be maintained.
This setting will vary from aircraft to aircraft and is denoted by pu . For the SR-22, this
value is about 20%. The engine is deduced to have failed, if power output is smaller than the
maximum power setting at which a successful power-on approach can be performed. This
value is referred to by pl . For the SR-22, this is about 5%. This value will again be different

35
 % power

Engine Status Detector

while status == “Indeterminate”!

if %power > p_u%!
status = “Engine Operational” !
elseif %power < p_l%!
status = “Engine Out”!
request pilot to increase throttle to cruise power!
else!
status = “Indeterminate”!
request pilot to increase throttle to cruise power!
end if;!
end while;!

engine status

Figure 2-2: Pseudocode for automatic engine status detection

for different aircraft. In case the power setting is in between pu and pl , the engine status
becomes ”Indeterminate”. This is a dangerous case, since it is not immediately handled by
the current system design without throttle adjustment from the operator. At this throttle
setting, the aircraft is neither capable of maintaining altitude nor capable of performing a
landing without risking nose-wheel barrowing or floating down the runway. Furthermore, the
shallow descent angle also makes the aircraft prone to hitting obstacles while on approach.
If this case is detected, autoland visually and audibly requests the pilot to increase throttle
to a cruise power setting, which is about 60%.

2.1.3 Landing Site Selection

The landing site selection function is responsible for choosing a landing site within range
that can be reached without colliding with terrain or known man-made obstacles along the
way. The term ”within range” requires that the achievable range is 30% higher than the
range required to reach the landing site from the current location in order to provide margin.
Landing site selection operates differently depending on engine status. The achievable range
is determined from glide performance in case the engine is out and from fuel range in case

36
the engine is operational. Furthermore, since there is a large difference in range between
cases engine out and engine operational, the landing site search is performed under more or
less demanding criteria levels. The criteria levels will be divided into levels 1-4.
In case engine operational, it is likely that there is a very large number of candidate
runways within range. Therefore, the landing site selection process can afford to be more
selective. However, a runway 1,000 miles away that is twice as long as a runway 50 miles
away, is not better than the runway close by, if the length of the runway close by is sufficient.
Hence, not all runways within range should be retrieved and compared to find the very best.
Rather, autoland searches for a runway that fulfills a set of ”good enough” criteria within
increasing search radii and stops the search process as soon as a runway is found that meets
the criteria. This ensures that runways closest to the aircraft’s current position are prior-
itized. For the engine operational case, the ”good enough” criteria simply consists of the
requirement that the runway is at least 5,000 ft long. This is the length required for rollout,
absorption of the touchdown point dispersion due to GPS/WAAS vertical position error,
error in estimation of wind and glide performance, and a 1,000 ft safety margin. Many other
criteria could have been applied, such as wind conditions at the landing site, runway surface,
availability of facilities, etc. However, runway length was determined to have the strongest
influence on success or failure of the landing and is therefore used as the single selection
criteria for the engine operational case. Any runway longer than 5,000 ft is considered to
fulfill criteria level 1.
If the engine is out, the range is dramatically reduced, which limits the number of can-
didate runways up to the point where no reachable runway can be found at all. Therefore,
it makes little sense to impose criteria level 1. Instead, autoland searches the full range for
a candidate runway and selects the longest one it could find. This runway is then said to
fulfill criteria level 2, which only requires that the selected landing site is a runway. In case
that no runway could be found, the landing site selector relaxes the criteria level further
and searches for a known emergency landing site, which are pre-loaded into the FMS from
a database such as the Worldwide Soaring Turnpoint Exchange maintained by the glider

37
community [11]. A known emergency landing site fulfills criteria level 3. Finally, criteria
level 4 consists of any unpopulated area.
In case the engine status is classified as ”Indeterminate”, the criteria level is initialized
to level 4. This is due to the aircraft not being able to perform a landing according to the
autoland approach procedure at this intermediate power setting. Hence, the goal is to guide
the aircraft away from regions were it could cause damage to the population and to minimize
the chances of the aircraft colliding with man-made obstacles.
Figure 2-3 illustrates the landing site selection procedure. First, autoland determines

trigger:
autoland mode set

calculate available
Range, R

Engine status = Engine Engine

"Indeterminate" Status? Operational

Engine Out

Initialize Criteria Initialize Criteria initialize Criteria

Level: CL = 4 Level: CL = 2 Level: CL = 1

prioritize landing sites

using search method no CL > 4 yes
for current CL

array of candidate landing site LS[]

initialize counter
CL = CL+1
c=0

calculate Initial
no c > length(LS)
Point for LS[c]

Initial Point

construct path to
Initial Point

length of path, L

L>R yes c = c+1

no

selected landing no feasible

site landing site found

Figure 2-3: Landing site selection process

38
the achievable range and initializes the criteria level. Landing site search is initialized with
criteria level 1 in case engine out, criteria level 2 in case engine operational and criteria level
4 for engine status indeterminate. The landing site prioritization function then searches for
candidate landing sites at the specified criteria level and orders them according to prefer-
ence. The operation of this function is detailed in the appendix, chapter A.1.3. For criteria
level 1 and 2, i.e. runways, candidates are ordered by runway length, with the approach
direction chosen as the headwind direction, if available. For criteria level 3, i.e. emergency
landing sites, distance to go to the landing site is used for prioritization. For criteria level
4, i.e. unpopulated area, the size of the area determines the position in the ordered list.
Autoland then steps through the list of candidate landing sites and attempts to generate an
obstacle-free trajectory to the landing site. If successful, the landing site is selected and all
lower ranked candidates discarded. If the end of the list is reached and no feasible trajectory
could have been generated, autoland relaxes the criteria level and repeats the search at the
new criteria level and the process repeats. Details of path planning function, that attempts
to generate an obstacle-free trajectory to the initial point are discussed in the appendix,
chapter A.2.4

2.1.4 Autoland Approach Procedure

After the landing site has been selected and a feasible path to the initial point determined,
autoland uses the trajectory following functionality of the the FMS to guide the aircraft
from its current position to the initial point. If there are no obstacles along the straight line
trajectory between the aircraft’s current location and the initial point, the aircraft flies a
straight line path to the initial point and aligns with the downwind leg by the time it reaches
the initial point as shown in Figure 2-4. In this illustration, it is assumed that the engine
is operational and that straight and level flight is possible. In case the engine is out, the
aircraft would lose altitude along the path to the initial point and the straight line trajectory
to the initial point would slope downwards.

39
 At the initial point, autoland requests power to be cut, if it is not zero already. While

Figure 2-4: Aircraft trajectory during autoland for case engine operational and no obstacles

waiting for the power cut to occur, autoland keeps the aircraft in a holding pattern at
constant altitude. The final approach trajectory is a standard right or left traffic pattern
consisting of downwind leg, base leg and final leg. After power has been cut, the aircraft’s
altitude, or expressed differently, the aircraft’s energy, could still be too high to start the
traffic pattern. In this case, autoland calculates the number and length of holding pattern
loops required in order to reduce the aircraft energy from its current level to the energy level
required to start the traffic pattern. Upon exit of the holding pattern, autoland adjusts the
length of the baseline traffic pattern to match the current aircraft energy and the expected
glide performance. Autoland begins to track the generated pattern, using the autopilot to
maintain the recommended approach speed of 90 kt.
As pointed out in chapter 1.4, the aircraft will only be able to follow the generated tra-
jectory at constant airspeed, if its performance matches the expected performance exactly.
Naturally, the actual glide performance of the aircraft will differ from the expected perfor-
mance. This difference will make the aircraft deviate off the trajectory, which would result
in the aircraft landing long or short of the target touchdown point. In order to correct any
error in expected and actual glide performance of the aircraft, autoland monitors the air-

40
craft’s energy status and adjusts the length of the trajectory in the lateral plane if deviations
off the planned trajectory in the vertical plane are observed. This scheme effectively makes
the path length the controlled variable to substitute for the lack of control over the aircraft
power. Figure 2-5 illustrates this process.
The autoland trajectory can be thought of to consist of three trajectory phases: The

Follow traffic
Baseline
pattern traj.
Initial
Fly straight
point Reduce Loiter at Initial Point Generate traffic trajectory Updated Energy
and level to
power to zero until E < E_max pattern trajectory trajectory error
Initial Point
Update traffic
pattern traj.

Figure 2-5: Sequence of functions performed during approach procedure in case engine op-
erational

trajectory to the initial point, the holding pattern and the traffic pattern. While in the hold-
ing pattern and traffic pattern, autoland uses trajectory updating to remove energy errors.
While the aircraft is tracking the trajectory to the initial point, autoland cannot shorten the
path since it already presents the shortest path found. Therefore, the trajectory is planned
including range margin, which is monitored as the aircraft proceeds towards the initial point.
If the range margin becomes depleted before the initial point is reached, autoland performs
re-planning of the currently selected destination and trajectory.

2.1.5 Baseline Holding Pattern and Traffic Pattern

The baseline holding pattern consists of two 180◦ , constant radius turns with zero straight
line segments in between the two semi-circles as shown in Figure 2-6. Elongation of the
holding pattern to compensate of energy errors is achieved by increasing the length of the
straight line segments. The zero length straight line segments were chosen as a baseline,
since this provides the smallest possible discretization of altitude levels, at which the holding
pattern can be exited.
The nominal bank angle, at which the DFC-90 autopilot performs turns to track the

41
trajectory is φ = 22.5◦ . At an airspeed of 90 kt, this results in a turn radius of 1,600 ft from
equation 2.1, which results in a circle track length of 10,000 ft.

V2
r= (2.1)
g · tan(φ)

where g = 9.81 sm2 . For a baseline glide ratio of 10:1, the altitude levels at which the turn
can be exited are spaced by 1,000 ft, which creates a challenge for the trajectory generation
process upon exit of the holding pattern. If the bank angle was increased to 30◦ , the turn
radius would reduce to 1,100 ft and the circle track length to 7,000 ft, which is a significant
improvement. More importantly, however, a small turn radius is needed while following the
traffic pattern, in order to allow trajectory updating for a sizable proportion of the overall
length of the traffic pattern. When using the current FMS trajectory tracking functions
without modification, trajectory updating is inhibited while the FMS captures the course
of a new leg. For the two 90◦ turns within the traffic pattern, this means that trajectory
updating is inhibited for one turn radius on the downwind leg and for two turn radii on the
base leg. Given that the usual length of the traffic pattern legs is between 3,000 - 5,000
ft, a turn radius of 1,600 ft for the smallest sized traffic pattern would completely inhibit
trajectory updating on the base leg. Even if the leg lengths were increased to 5,000 ft, the
proportion of the traffic pattern available for trajectory updating is still unsatisfactory at
less than 50%. Therefore, it is recommended to increase the nominal bank angle during
turns to 30◦ , which will be assumed to be the nominal bank angle for the remainder of this
thesis.
The baseline traffic pattern trajectory is a standard, rectangular, right or left traffic pat-
tern as shown in Figure 2-6. The traffic pattern direction is chosen so that the runway is
not crossed during the aircraft’s approach to the initial point. Implementation details are
provided in the appendix, chapter A.1.4. The length of the pattern legs is a tradeoff between
the capability to correct energy errors as the aircraft follows the downwind leg and base leg
and the amount of energy error accumulated on the final leg. The shorter the final leg, the
less energy error can be accumulated on the final leg, but at the same time, the less energy

42
 L_F = 1,500m = 6,000 ft
r
F_F

Target touchdown point, TD 1000 ft 500 ft r

Latest intersection
of updated base leg

Margin for potential

Margin for aircraft to intercept
undershoot due to
final leg from base leg L_B = 1,500 m Lloop 2rπ
glide performance E
= 5,000 ft loop
= =
estimation error and L/Dl L/Dl
GPS error
L/Dl = L/D0 · cos(φ)3/2 = 0.8 · L/D0
φ = 30◦
r=340 m = 1,100 ft
where
V=90 kts
L/Dx represents the (L/D)expected on the leg denoted by sub
Initial point, IP Excess energy ∆E can be found r from:

∆E/W = E/W − ET D /W − Ebase /W

r F_B
The excess energy ∆E is allocated by moving F D outwards by dis
L_D=1,500 m = 5,000 ft
r distance to be added to the minimum length of the downwind leg and fin

Holding pattern ∆E/W

∆L =
(L/D)−1 −1
d + (L/D)f

Figure 2-6: Length and shape of the

where baseline
(L/D) trajectory
d and (L/D) f are the(L/D)expected on the downwind leg an
Hence, the total lengths between the fixes become:

L�D = LD + ∆L : distance between Initial Point and down

error can be corrected while still on the base leg. TheL�Bmaximum
= LB amount, bybetween
: distance which downwind
the leg fix F D
L�F = LF + ∆L : distance between base leg fix F B and ta
remaining trajectory can be shortened at the start of the base leg is 30%, which is achieved,
The position of the unknown downwind leg fix F D and base leg fix
if base leg and final leg are of equal length. A knowledge
similar argument
of the totalapplies to and
leg lengths the the
downwind
location ofleg
the initial point, I a
Using I as the origin of a local North-East coordinate system, the locatio
and base leg, which motivated the use of all equal length legs. The length of the legs was
F DN orth = LD cos(χD )
chosen to be 5,000 ft in order to provide a trajectory updating capability of greater than
F DEast = LD cos(χD )
50% of the time on the downwind leg and base leg if the turn radius is 1,100 ft. The choice
Similarly, using T as the origin of a local North-East coordinate syste
of 5,000 ft of base leg length presents a compromise between trajectory updating
F B
capability
= L cos(π − χ N orth D F)

on the base leg and the goal to minimize the length of the final leg. F BEast = LD cos(π − χF )
Trajectory updating capability during the turns
π − χFwould allowtheforcourse
represents continuous energydirection
in the opposite correc-of the final leg cou
tion capability and allow for a shorter traffic pattern and hence
7.12 Energy final Manager
Curve leg, but would require
the FMS to accept course changes, while it is capturing a previously updated course. Alter-
7.12.1 Inputs

natively, a direct interface to the autopilot could• be

L F: length of straight
implemented. Forlinethe
segment of final
current leg excluding
design, the turn
margin
the simple autoland/FMS interface is deemed sufficient. For subsequent iterations, perfor-
• L B: length of straight line segment of base leg excluding turn segm
mance enhancements can be achieved by modifying the
• L D:autoland/FMS
length of straight interface if considered
line segment of downwind leg excluding the
• (L/D)expected : expected glide ratio, one per segment of the trajecto
43
• E TD/W: desired touchdown energy
• r: aircraft nominal turn radius

37
necessary.

2.1.6 Trajectory Planning using Energy Management

In order to determine the number and length of loops the aircraft follows in the holding
pattern and to determine the length of the traffic pattern after the aircraft has exited the
holding pattern, autoland requires knowledge of:

1. The amount of energy available to dissipate normalized by the aircraft weight ∆E/W .

2. The rate of energy dissipation with ground distance travelled.

The aircraft energy at any point in time is defined as the sum of kinetic and potential energy.

1
E = mgh + mV 2 (2.2)
2

By normalizing the energy, E, by the aircraft weight, W , no knowledge of the exact aircraft
mass is required to compute the numerical value of E/W .

V2
E/W = h + (2.3)
2g

Furthermore, the normalization by weight results in E/W being expressed in units of altitude,
which allows ”energy” to be replaced by ”altitude” whenever the airspeed does not change.
The amount of energy available to dissipate, ∆E/W , is simply the difference between the
current energy, E/W , and the desired energy at touchdown, Etd /W .

V02 V2
∆E/W = E/W − Etd /W = (h0 + ) − (htd + td ) (2.4)
2g 2g

where h0 and V0 denote the current altitude and airspeed respectively, while htd and Vtd
denote the touchdown zone elevation and desired airspeed at touchdown respectively.
While the aircraft follows the pattern, the airspeed is set to the best glide speed, Vbg ,

44
which is 90 kt for the Cirrus SR-22. In order to reduce the descent rate before touchdown,
autoland performs a flare maneuver that causes the airspeed to reduce from Vbg =90 kt
to Vtd =75 kt. Hence, if V0 = Vbg , the energy available to dissipate simply becomes the
difference between the altitude of the current location and the altitude of the touchdown
zone, plus the difference in kinetic energy resulting from reducing the airspeed from 90 kt to
75 kt, which is equivalent to an additional 100 ft of altitude.
The rate of energy dissipation with ground distance travelled, R, can be derived by
differentiating equation 2.3 with respect to R and setting the airspeed, V , equal to the
constant best glide speed, Vbg . This is justified even though the airspeed is changing from
Vbg to Vtd during the flare phase, since the flare is an out-of-trim maneuver, where the reduced
altitude loss is compensated for by a reduction in airspeed. The total loss of energy over
ground distance travelled during the flare maneuver, however, is the same as during the
gliding approach phase. The energy dissipation rate with ground distance travelled can be
written as:
d dh 1
(E/W ) = = (2.5)
dR dR L/D
Due to the trajectory being planned for constant airspeed, the rate of change of energy
normalized by weight is only dependent on the rate of altitude lost over ground distance
travelled. This value is known as the aircraft’s L/D or glide ratio. It is directly related to
1
the flight path angle γ by the relation: tan γ = L/D
. Here, it is assumed that the glide ratio
and flight path angle are referenced to the inertial frame. If no wind is present, the glide
ratio of the aircraft as measured with respect to the airmass and as measured with respect to
the ground is equal. If wind is present, the aircraft travels a different distance over ground
than it travels through the airmass. The convention adopted here, is that the ”glide ratio
through the airmass” is adjusted for wind effects so that glide ratio always refers to the ratio
of altitude lost, over ground distance travelled.
From knowledge of the wind speed and direction at the selected landing site, the course
of the downwind leg, base leg and final leg, as well as the nominal bank angle during turns,
autoland calculates the expected glide ratio on each segment of the holding pattern and the

45
traffic pattern. The segments of the holding pattern are: Turn to outbound leg, outbound leg,
turn to inbound leg and inbound leg. The segments of the traffic pattern are: Downwind leg,
Turn 1, Base leg, Turn 2 and Final leg, where Turn 1 represents the turn from the downwind
leg to the base leg and Turn 2 represents the turn from base leg to final leg. The location of
each leg is illustrated in Figure 2-7. The expected glide ratios on each of the traffic pattern

final leg
turn 2

base leg

turn 1
inbound leg downwind leg

turn to inbound leg turn to outbound leg

outbound leg

Figure 2-7: Naming convention for holding pattern legs and traffic pattern legs

segments are referred to by L/Dd , L/Dt1 , L/Db , L/Dt2 and L/Df in this order. The energy
dissipated on the base leg segment for example becomes:

1
∆Eb /W = · LB (2.6)
L/Db

where LB denotes the length of the base leg segment. The energy dissipated on the remaining
segments can be calculated in a similar manner. Assuming constant glide ratio along each
individual segment, energy dissipation along each segment is linear with ground distance
travelled.
The complete ”energy vs. range-to-go until the planned touchdown point” curve for the
rectangular traffic pattern therefore becomes a piecewise linear, continuous function as shown
in Figure 2-8. The currently available energy E/W at the current range to go until the target

46
 touchdown point, R, is shown on the right hand side of the the figure. The desired energy
at touchdown, Etd /W , with zero range to go, is shown on the left hand side. The required
length of the traffic pattern is found from matching the length of the downwind leg, base
leg and final leg to the energy available to dissipate, ∆E/W = E/W − Etd /W . The length
of the turn segments are a result of the choice of the traffic pattern shape andVlength of the −t
v,cmd = (Vv,0 − 3.3) · (1 − e ) + 3.3

pattern legs and cannot be chosen freely for energy matching.

Vv,0 = V · tan(γ)
In the holding pattern, the length of the pattern is adjusted to match the energy
γ = arcsin(L/D)

h − htd
R =Energy
0.77 · currently
available, L/D
E/W

Vtd2
Current position:
Etd /W = htd
On final leg+ 2g
2
V
E/W = h +
2g
χc2
Vv,cmd = (Vv,0 − 3.3) · (1 − e−t ) + 3.3 (20) χc1
αp = 15◦
αp = 45◦
◦
Vv,0 = V · tan(γ) (21) αp = 75
−−→ −→
dtotal =| FB,0 , FB |
γ = arcsin(L/D) (22)
2 Design Concept
h − htd
REnergy on ·
= 0.77 Length of(23)
base leg, L_B
-Engine on guidance to Initial Point -Engine off approach from Initial Po
L/D
touchdown, E_td/W
-Use Holding pattern -Use trajectory
27 updating for energy management
Vtd2
Etd /W = htd + (24)
3 Functional requirements
2g

V2 The autoland function is designed to allow an aircraft to semi-autonomously

E/W = h + (25)
throttle commands) perform an emergency landing from any point of initiat
2g autolanddistance
needs to employ several functions.
Figure 2-8: Energy dissipation with ground travelled
1. Select destination
2. Guide the aircraft to an initial point on the standard traffic pattern a
difference between the current energy level E/W and the energy required to fly the traffic
3. Hold the aircraft on a holding pattern by the initial point until the
pattern E/Wtp,0 . E/Wtp,0 can be derived from knowledge within
of the length
a band of the
of energy traffic
required patternthe standard traffic pa
to commence
−−→ −→ 4. Guide the aircraft along the traffic pattern while updating the traffic
dtotal =| F
segments B,0 , F
and B | expected glide performance on each
the (26) segment:
energy deviates from the expected energy

5. Perform a flare maneuver as the aircraft reaches the flare altitude ab

cept LD − r rπ LB − 2r6. Capability
rπ to absorb − r in wind measurements and uncertainty in
LF errors
E/Wtp,0 = + + + + (2.7)
L/D
Initial Point -Engine off approach from Initial Point -Use 2L/Dt1
d standard L/Db 7. Capability
traffic pattern 2L/Dt2to absorb
L/D f in L/D estimation
errors 0
se trajectory updating for energy management
The length of the holding pattern is adjusted by shortening or elongating the length of the
requirements
inbound leg and outbound leg and number of loops in the pattern. Details on the adjustment6
designed to allow an aircraft to semi-autonomously (with the pilot following simple
orm an emergency landing from any point of initiation at altitude. For this
47 purpose,
y several functions.

to an initial point on the standard traffic pattern at the selected destination

on a holding pattern by the initial point until the aircraft’s energy has reduced to
of the length of the holding pattern and traffic pattern to match energy targets are discussed
in chapter 2.1.9.

2.1.7 Performance Prediction using Online Performance Estima-

tion

In order to determine the energy dissipation with distance over ground flown for each traffic
pattern segment, autoland requires information on the expected glide ratio of the aircraft
on each segment. The Cirrus SR-22 manual states that the glide ratio at best glide speed
is 9.6:1 with flaps retracted, which could potentially be used as an input for performance
prediction. However, there is no guarantee that the aircraft will be in a clean configuration
when autoland is initiated. Furthermore, the propeller could be windmilling or be stopped,
which would make a large difference in observed glide ratio. Therefore, autoland monitors
the aircraft’s performance as the aircraft glides at best glide speed and uses the deduced
glide ratio for performance prediction on the following traffic pattern legs. For this purpose,
autoland first determines an instantaneous glide ratio, L/Di from the change in energy,
∆E/W , and ground distance travelled, R, within one sampling interval. Autoland deduces
the change in energy from the change in altitude and airspeed, and deduces the distance
travelled from the change in position as shown in Figure 2-9.
This instantaneous glide ratio contains the effect of wind and any non-zero bank angle,
which both modify the measured glide ratio from the glide ratio that would be observed
in still air with wings level. However, this no-wind, wings-level glide ratio is required for
performance prediction on subsequent legs. Hence, any wind and bank angle effects are
removed from the measured, instantaneous glide ratio, L/Di , to yield L/Di,0 . The wind
effect is broken up in crosswind and tailwind effect, with the tailwind having the strongest
influence on the glide ratio. A 25% tailwind results in a 25% increase in glide ratio. The
crosswind effect is much smaller but still yields a change in glide ratio of 3% for a 25%
crosswind. Since glide ratio estimates are expected to be accurate to within 5%, this effect

48
 • L/D0 : aircraft no-wind, wings-level glide ratio. (Note: vertical wind effects
non-zero Ivertical
north = wind + (LD −inLthis
Pnorthincluded F ) cos(π − χd )
number)
Chapter 1 Ieast = Ieast + (LD − LF ) sin(π − χd )
7.8.3 Trigger
Sampling clock, every 1 s. ISampling time selected to balance error from using stra
north = L cos(χ)
when following arcs and robustness to noise. Using airspeed = 44 m/s and bank a
Introduction
!R
traverses θ = 5.3◦ of the full
hi
Ieast = L in
circle sin(χ)
Vi 1 s. The % error made in measuring the
Pi+1 calculated from the following formula:
Course to initial point:
!h
hi+1
χ i = atan2(∆E, ∆N )
�
Vi+1 xarc − xline r 2(1 − cos θ)
%error = = < 0.1%
Pi Autoland systems � xarc rθ
∆R = ∆N 2 + ∆E 2
!V
7.8.4 Description
2
Vratio
1.1 Motivation forThis Design of Autoland
module measures Ei /W =System
an apparent glide
h i+
i (L/D)observed and corrects this value
!R effects to arrive at the no-wind, wings-level 2g glide ratio (L/D)0 . The apparent glid
the change in energy, ∆E/W = Ei /W −Ei−1 /W , and distance travelled over groun
Accident statistics from [1, ?]
The instant apparent glide ratio becomes the ratio of distance travelled over grou
the sampling interval. These instant measurements are then smoothed and averag
average filter∆R
as shown in Figure 23
L/Di = (1.1) 2
∆(E/W ) Vtas,i
Ei = hi +
30 2g
Accident statistic for Engine Out: Referene: [2].
Vw , !w " ϕ
(wind speed for General Aviation
Accident statistics (current vs Commercial aircraft.
(currentLittle safety en- 29
and direction) heading) bank angle)
hancing technology in General Aviation compared to Commercial Aircraft. However,
Tailwind Crosswind Bank angle L/Di,0
L/Di
need is there correction
as accident statistics show. correction correction

Technology opportunities. Digital Autopilots, GPS/WAAS position sensing. Elec-

! (current course) !
trical servos.

L/Di,0 L/Di-9,0 L/D

1.2 Design Challenges - Problem Statementi-10,0
where γ = π− | (χ1 − χ2 ) |.
Autoland capability long proven on commercial
�n aircraft. What makes it hard on GA
(L/D) i,0
aircraft? L/D0 = i=1 (A.58)
n
1) No auto-throttle 2) No radar altimeter 3) No rudder 4) No auto-brakes
Requirements from Stakeholder: 1) No hardware upgrades allowed
Figure 2-9: Determination of observed glide ratio while in flight
15

needs to be included. In addition to wind, the glide ratio is affected by the aircraft’s bank
angle during a turn. For any non-zero bank angle, the aircraft’s glide performance is reduced,
since a part of the lift vector is now used to keep the aircraft in the circular motion. For a
30◦ bank angle, the glide ratio reduces by 20%, which clearly is a non-negligible effect.
Since the measurement of L/Di , wind and bank angle are assumed to be noisy, autoland
uses a simple moving average filter to smooth the samples of L/Di,0 to arrive at the final
output L/D0 . This L/D0 can now be used to predict the performance of the aircraft on
each of the subsequent pattern legs, L/Dexp . From knowledge of the course of the leg to be

49
flown, χ and the wind speed and direction, the crosswind and tailwind components can be
calculated and their effect included from knowledge of the aircraft’s nominal airspeed Va and
sink rate Vsink as shown in Figure 2-10. For a detailed description of the equations involved
in this algorithm, refer to the appendix, chapter A.2.2 and A.2.3.

ϕnominal Vw , !w Vw , !w

L/D0 Bank angle Crosswind Tailwind L/Dexpected

correction correction correction

Va , ! Vsink , !

Figure 2-10: Calculation of the expected glide ratio, L/De xp, from L/D0

2.1.8 Aircraft Energy Monitoring and Energy Error Detection

As the aircraft is following the traffic pattern, autoland compares the aircraft’s energy and
range to go until the planned touchdown point with the stored energy vs. range to go curve
as was presented in Figure 2-8. If autoland determines that the energy is higher than the
reference energy, the trajectory will need to be elongated to match the available excess en-
ergy. If the energy is lower, the trajectory will need to be shortened. Figure 2-11 illustrates
this concept for the case that the aircraft has excess energy on the downwind leg. The red
and blue boundaries are the excess energy and lack of energy tolerance bounds respectively,
which are chosen in such a way that the aircraft is allowed to have 5% more energy than
needed to reach the start of turn 2 and 3% less energy than need to reach the turn. These
values were chosen as a tradeoff between noise tolerance and the desire to update as often as
the FMS allows, to keep energy errors as small as possible. Noise added to the measurement
of aircraft energy and range-to-go should not cause the measurement to cross the tolerance
bounds and cause an update. Preliminary analysis shows that tolerance bounds of +3% and

50
 E/W E/W
current aircraft
energy, E/W

+ dE/W

Range Range
to go Elongated Elongated Elongated to go
Final leg Turn 2 Base leg Turn 1 Downwind leg Final leg Turn 2
unchanged Base leg turn 1 due downwind
unchanged
to increase leg
in course
change

Figure 2-11: Concept of excess energy correction through trajectory elongation

-3% meet this requirement. In addition to that, the operation of the FMS limits the fre-
quency of trajectory updates to greater than every 5 s. For expected L/D estimation errors
on the order of 10%, this consideration can be met by tolerance bounds of +2% and -1%.
Hence, the tolerance bounds imposed by measurement noise become the active constraint.
Since, there is little concern about the capability to correct a positive energy error, the upper
bound was increased from the minimum of 3% to 5% to improve performance in a more noisy
environment. For details on how the energy vs. range curve-to-go curve and its boundaries
are stores is discussed in detail in the appendix, chapter A.2.9.
Using the trajectory updating strategies as outlined in chapter 2.1.9, the legs of the tra-
jectory are elongated or shortened so as to match the available aircraft energy to the energy
expected to be dissipated along the updated trajectory. The reason for the aircraft deviating
off the expected energy vs. range to go curve is either due to the actual aircraft performance
differing from the expected performance, or due to measurement errors of wind speed or
direction, Vw and χw , or aircraft glide performance, L/D0 .
In order to generate a new trajectory that matches the actual aircraft performance, up-
dated performance estimates will need to be used and the slopes of the reference energy
vs. range-to-go curve adjusted to the actual performance observed. Therefore, autoland
continuously runs the aircraft performance estimator and updates the reference energy vs.

51
range-to-go curve as soon as new estimates are received. This concept is illustrated in Figure
2-12. In this case, the aircraft performance estimator determined that the actual L/D0 is
lower than previously expected and the wind speed and direction has changed. This example
is certainly exaggerated, but clearly shows the consequence. If the aircraft’s energy was right
on the expected energy vs. range-to-go curve before the update, the aircraft’s energy would
be well below the energy vs. range curve after the update.
This is one mechanism through which a trajectory update is triggered. The other mech-

Initial estimate: Updated estimate:

(L/D)0 = 9.6:1 (L/D)0 = 8:1
Vw = 10 kts @180° Vw = 20 kts @90°

Figure 2-12: Update of energy vs. range-to-go reference curve upon receipt of new aircraft
glide performance and wind estimate

anism is the aircraft falling below the energy vs. range curve without a difference in actual
and expected performance having been detected. This scenario is possible for example, if
the wind speed and direction from the wind estimator is in error.

52
2.1.9 Trajectory Updating

Trajectory updating in order to remove energy errors from an overestimation or underesti-

mation of the expected glide ratio, L/Dexp , is performed while in the holding pattern, while
on the downwind leg and while on the base leg. The current design does not provide energy
error correction capability on the final leg, but absorbs energy errors on the final leg by ap-
propriate placement of the target touchdown point. Since the autoland/FMS interface does
not allow for trajectory updating while the FMS is capturing a previously updated course,
autoland suspends trajectory updating for the duration of the course capture. This includes
the instances where the FMS is capturing a new course after an unscheduled trajectory up-
date and the scheduled transitions from downwind to base leg in turn 1 and base leg to final
leg in turn2.

Holding Pattern

The holding pattern flown while waiting for the power to be cut or while depleting excess
energy consists of two half turns with variable length straight line segments in between.
The goal of this function is to bring the aircraft to exactly the energy state required to fly
the baseline length traffic pattern, E/Wtp,0 . In order to achieve this, autoland adjusts the
number of the turns and the length of the straight line segments to match the energy to be
depleted, ∆E/W = E/W − E/Wtp,0 − E/Wtd , where E/W is the current aircraft energy.
With this method, the length of the straight line segments will range from 0 to equal the
length of a half turn. The holding pattern direction will be either right or left depending on
the direction of the traffic pattern used for the final approach. The convention used here is
that the aircraft first aligns with the course of the downwind leg at the initial point and then
proceeds to turn away from the traffic pattern into the holding pattern as shown in Figure
2-13 for a left traffic pattern.
With this convention, the aircraft would perform right turns in the hold for a left
traffic pattern and left turns in the hold for a right traffic pattern. The requirement to first

53
 L

Initial Point

basic holding pattern

with zero length
straight line segments

Holding pattern with

elongated straight line
segments

Figure 2-13: Direction and sizing of holding pattern

align with the downwind leg at the initial point is included to allow direct transition to the
traffic pattern if the energy by the time of arrival to the initial point has dropped below
the maximum allowable energy to start the traffic pattern as shown in the state transition
diagrams in chapter 3.2.2. Due to the fact that the aircraft cannot deplete less than the
energy required for one 360◦ turn with zero length straight line segments, E/Wloop , there
must be a range of energies at which the traffic pattern may be started. E/W can be
calculated from knowledge of the turn radius, r, and expected glide ratio within the turn
L/Dt . For a baseline glide ratio of 10:1, a 30◦ bank angle during the turn reduces the glide
ratio by 20% to 8:1. Hence, the energy dissipated during one full turn becomes:

2rπ 7, 000f t
E/Wloop = = = 875f t (2.8)
L/Dt 8

To spread the distribution of energies at which the traffic pattern is started equally in
both directions, the maximum allowable energy to start the traffic pattern is chosen as

54
 1
E/Wtp,max = E/Wtp,0 + 2
· E/Wloop + E/Wtd . With this choice, the minimum energy at
1
which the traffic pattern is started becomes E/Wtp,min = E/Wtp,0 − 2
· E/Wloop + E/Wtd .
Hence, if the aircraft’s energy is just above the E/Wtp,max threshold on alignment with the
downwind leg at the initial point, it will enter the holding pattern and fly one loop with zero
length straight line segments. On passing the initial point the next time, its energy would
have dropped to just above E/Wtp,min and it will exit the pattern. If the aircraft’s current
energy is higher than E/W = E/Wtp,0 + E/Wloop + E/Wtd , autoland adjusts the number of
loops and length of the straight line segments to bring the aircraft to an energy state equal to
E/Wtp,0 by the time of passing the initial point after an integer number of loops according to
the following method: The number of loops to be performed after completion of the current
loop, n, is found from the integer number of loops with zero straight line segments that can
be performed with the current energy level E/W .

E/W − E/Wtp,0 − E/Wtd − E/Wip

n=� � (2.9)
E/Wloop

where �x� is the greatest integer y such that y ≤ x (rounding down). The term E/Wip is
the energy required to complete the current loop, which can be found from knowledge of
the current location within the loop and the expected glide performance on the straight line
segments and the turns.
Due to the discretization of energy levels by steps of magnitude E/Wloop , there may be
an excess of energy ∆E/W of magnitude [0, E/Wloop [. This excess of energy is allocated to
each loop by equal amounts and absorbed with equal amounts by each loop by elongating
the straight line segments. ∆E/W can be written as:

∆E/W = E/W − E/Wtp,0 − E/Wip − E/Wtd − n · E/Wloop (2.10)

The length of each straight line segment, L, then becomes:

∆E/W
L= (2.11)
n(L/Dd + L/Df )

55
 d0
L/Dav
d = d0
10
χf
αp = 30◦ χf
αp (2) αp = 30◦
α αp (2)
where L/Dd and L/Df are the glide ratios expected on the inbound leg (equal
α to the course
LB (L/D)b + (LF I + LF )(L/D)f
L/Dav =
of the downwind leg) and the outbound leg (equal to the course of the
LB final
+ LF I leg).
+ LF As soon L/Dav =
as n = 0, autoland waits until the aircraft has turned onto the inbound leg
αp = χc2 − χand
b signals the
FMS to exit the hold. This causes the FMS to proceed to the
�
downwind leg assuming it is
FB,north = d cos(χc1 ) + FB,north
�
already programmed. FB,n
�
FB,east = d sin(χc1 ) + FB,east
FB�
�
FF,north = d cos(45◦ ) cos(χc1 ) + FF,north
Downwind Leg 1 Equations �
FF,north
1 Equations
�
FF,east = d cos(45◦ ) sin(χc1 ) + FF,east d=
d = f (∆E/W, L, αp , L/Dav ) � f (∆
FF,east
While on the downwind leg, autoland
χc
elongates and shortens the trajectory
χc as shown in Fig-
FF� I,north = d cos(χ ) + FF I,north
αp (1)the planned trajectory, whereas theαred
ure 2-14. The blue curve represents p (1)
c2
curve represents FF� I,n
αp (2) αp (2)
� α
FF I,east = d sin(χc2 ) + FF I,east
the trajectory after the update.α
FF� I
LD (L/D)d + LB (L/D)b L
In case of excess energy, autoland moves the base leg fix, L/Dav =
FF
L/D� avB,
= outwards
F,north = LI cos(χ along
L f+) L
the course
+ FF,north
D B �
FF,n
�
FF,east χc −f χ
=αpLI=sin(χ ) + FF,east
F_F F_F’
1 Equations F_F FF�
F_F’ �
�
FB,north = atan2(∆E,
d cos(χc ) + ∆N
FB,north FB,north
χ� = )
d =f if (∆E/W, L, αp , L/Dav )
Excess energy: 1 Equations
Lack of energy: χc2 d �positive= d sin(χ ) + F
->F
�
FB,east
χc 1 1 Equations
-> d negative
χc1 Equations B,east cχ
c2
B,east
αp (1) ◦
F_B’ αp = 15
αp (2) ◦ d = f (∆E/W, L, αp , L/Dav )χc1 d(1)
= f (∆E/W
αχp = 45 χd = atan2(∆E, α χ= c 15◦
χc p ∆N ) ◦
χd =
αc d α
αp =αp75◦ χd α =
αpp F_B45p (1)
χd α
α F_B
LD (L/D) αα +
dp−
= 75p (2)
◦
d LB (L/D)b
L/Davd= −− → →
L B,0 , FB α F_B’
L total =| F LD + L|B
2 2Design DesignConcept
Concept 2 Design 2 Design Concept
Concept
αp = χc − χ
2 -Engine
-EngineDesign
on guidance Concept
on guidance to Initial
to InitialPoint
Point-Engine off-Engine
off approach
-Engine approach onfrom
from -Engine
Initialguidance
Point -Use to
on guidance toInitial
Initial
standard
Initial Point Point
Point
traffic
-Use -Engine
-Engine
pattern
standard off tra
apo
-Use Holding pattern -Use trajectory updating for energy 2 -Use Design
management
-Use
-Use Holding pattern -Use trajectory updating for energy management Holding Holding pattern
pattern Concept
-Usetrajectory
-Use trajectory updating
updatingfor e
Figure 2-14: Trajectory
-Engine
2 Design on updating
guidance to method
ConceptInitial Pointon -Engine
the downwind
off approach legfrom Initial Point -Use standard tr
-Use 3Holding pattern -Use
Functional requirementstrajectory updating for -Engine
energy on guidancerequirements
3 management
Functional to Initial Point -Eng
3 -Engine
Functional requirements 3 Functional
-Use Holding
on guidance to Initial Point -Engine off approach from Initial Point requirements
pattern -Use -Use
trajectory
standardupda t
◦
2
The autoland
The autoland function is designed to allow an aircraft to semi-autonomously
-Use Holding pattern -Use trajectory updating for energy management
Design
function
(with the pilot Concept
is designed to simple
following allow an aircra
χc1 , which is angled at 45 toThe
theautoland
course
throttle of theperform
commands)
function baseline
is designedleg
an emergencyand downwind
landing
to allow an
fromThe
to leg,
any pointthrottle as shown.
commands)
of initiation
aircraftautoland
at altitude.
function
semi-autonomously At
perform
Foran emergency
this
is designed purpose,
(with theto allowlandingaf
pilot an
follo
autoland needs to employ several functions. autoland needs on
-Engine to employ severaltofunctions.
guidance Initial Point
throttle commands) perform an emergency landing from any point of initiation at altitude. For tthrottle
5 commands) perform an emergency land
the same time, autoland moves3the 1.Functional
autoland
final
needs
leg fix, F F, away from the runway
to employ requirements
Select destination several functions. autolandthreshold
-Use
SelectHolding
1.needs to in such
employ pattern
destination several-Use trajectory
functions.
a way that the course of the base leg
1. Select
The is
2. Guide maintained. Similarly,
the aircraft to an initial
destination
autoland function is designed
point on thefor lack
standard of
traffic
1.
to allow an aircraft toenergy,
pattern
Select at theautoland
2. Guide aircraft
selected to
destination
semi-autonomously
an initial point on the st
destination
(with the pilot fol
throttle commands) perform an emergency landing2.from anythe
point of initiation 3. Hold
3. Hold the aircraft on a holding pattern by the initial point until
3
the aircraft’s
at altitude. For
Functional requirem
the aircraft on ahas
energy holding pattern
reduced to by the
moves F B inwards along χc1 by2. the amount
Guide banddofand
thea aircraft moves
to an initialF F towards
topoint on the Guide
the
standardrunway
traffic
within aircraft
band of to
threshold
apattern atanthe
energy initial point
selected
required on t
destin
to commence
autolandwithin
needs energy
to employ required
several commence
functions. the standard traffic pattern
The autoland function is designed to
4. Guide
the the aircrafton
along the
◦ traffic pattern
pattern while 3.initial
Hold the aircraft on aircraft’s
aasholding pattern by
4. Guide the aircraft along the traffic pattern wh
along the course of the final leg3.by
Hold
amount
1. Select
aircraft
d · cos(45
energydestination
a holding
). The shiftby
deviates from the expected energy
ofupdating
the
the the throttle
base traffic
point
leg
withinenergy
pattern
until length
parallel
deviates
a band
the
fromto
commands)
of energy
the aircraft’s
perform
therequired
energy has
an emer
expected energy
to comm
within a band of energy required to commence the standard autoland trafficneedspattern to employ several fun
the baseline base leg reduces the amount
5. Perform athat
flare F B
maneuver is
as required
the aircraft to
reaches move
the flare and 5.
altitudePerform
hence
above a
the flare
reduces maneuver
landing site as the aircraft reach
2. Guide the aircraft to an initial point
4. Guide the aircraft along the traffic pattern while updating on the standard
4. Guide traffic
the pattern
aircraft
the traffic along at thethe selected
pattern traffic
lengthpatter dest
as t
1. Select destination
6.
energy Capability
deviates to absorb
from errors
the in wind
expected measurements
energy and energy
6.
uncertainty deviates
Capability
in wind to from
absorb
profile withthe
errors expected
in
altitude wind energy
measurem
the course change from the downwind
3. Holdleg the to the base
aircraft leg compared
on a holding pattern bytothe a scheme
initial point where
2.a Guide
until only the aircraft’s energy h
the pattern
aircraft to an initialr
7. Capability
within
5. Perform
to absorb
aa flare
band errors inrequired
of energy
maneuver
L/D estimation
to commence
as the0 aircraft reaches the7.flare
5. Perform
the
Capability
standard flareto
altitude trafficabsorb
maneuver
above the
errors in L/D
as landing
the 0 estimatio
aircraft
site
8. Capability to absorb up to of 30% error in an initial (L/D)expected 8. Capability
while
3. to absorb
the aircraft
Hold the upthe
to of
isaircraft
on 30%
downwind
on error
a in an in
holding
6. 4.Capability
Guide
leg andthe
base 56if it takes
toaircraft
absorb
leg along10 sthe
errors in wind
to accumulate a stable,6.
trafficmeasurements
pattern newCapability
while updating
and
L/D andto
leguncertainty
estimate base absorb
the
(drives legtraffic
ifinit errors
ofpattern
wind
takes in
s towind
10profile lengthmeasu
with
accumulate as
ala
energy deviates within achoice
band leg
of lengths).
energyrange requir
This number was from
derivedthefromexpected
the range of energy
L/D measurements This takennumber
during was derived from
an Avidyne flightthetest, of L
7. Capability to absorb
simply using VT as /Vsinkerrors in L/D
. Applying estimation
wind0corrections 7. Capability
should improvesimply to absorb
using
the accuracy VT asand errors
/Vsink inshould
. Applying
averaging L/D 0 estim
wind correc
5. Perform
get rid ofatheflare
highmaneuver as the
frequency spikes. aircraft
Hence +/-30% reaches theget
is considered
4.beGuide
flare
to rid altitude
aof the hightheabove
pessimistic
aircraft
frequency
estimate. the along
landing
spikes. thesite
Hence tr
+/
8. Capability to absorb up to of 30% error in an initial 8. Capability
(L/D)expected to absorb
energy upthe
deviates
while to aircraft
of
from30%the erroron in
is expecth
9. Capability tofinal
absorb 15% error in (L/D)
6.leg9. Capability to
Capability
and base leg toabsorb
absorb 15%errors
if it takes error inin
10 s (L/D)
towind after having
measurements
expected
accumulate leg intercepted
a stable, and
and
new base the
leg
uncertainty
L/D if itlegtakes
estimate in(drives
wind 10choice
(drives sprofileof
tochoice
accumul
expect
with
of l
target touchdown point). This choice is a compromise between target
5.
runway touchdown
Perform
consumed apoint).
due flare
to This choice
maneuver
placement of isrange
a com
as th
Thisthe number was derived from the range of L/D This number
measurements
the target
was derived
taken
touchdown during
point
from
and an theAvidyne
tolerance to
7.simply target
Capability touchdown
to absorb point and
errors tolerance
in L/D to errors
estimationin (L/D) expected originating from errors in wind
using V /V . Applying wind 0 corrections simplyshouldusing VT asuncertainty
improve
measurement, /Vthe . Applying
sinkaccuracyin wind and andwind
avera
error c
in
the base lex fix, F B is moved.
The value of d for a given energy error ∆E/W , can be found from knowledge of the
distance to the base leg fix, L, the angle that χc1 is making with the course of the downwind
leg, αp , and the average glide ratio expected on the remaining legs, L/Dav . αp will be 45◦ for
the first update but will be different to this value after the first update. ∆E/W , L and αp
affect d non-linearly and in such a way that an explicit expression for d in terms of the former
parameters cannot be found. Therefore, autoland uses lookup tables as described in more
detail in the following section to retrieve d from a given ∆E/W , L and αp using interpolation
between pre-computed values. L/Dav , however, acts as a simple linear scaling factor, which
will be used to adjust the output from the lookup table. In summary, in order to find d,
autoland first retrieves d0 , which specifies the magnitude and direction by which F B would
have to be moved if L/Dav = 10, using the lookup tables and subsequently adjusts d0 for
the actual L/Dav to arrive at d:

d0 = f (∆E/W, L, αp ) (2.12)

with
αp = χc1 − χd (2.13)

Then,
L/Dav
d = d0 · (2.14)
10
with
L(L/D)d + LB (L/D)b + LF (L/D)f
L/Dav = (2.15)
L + LB + LF

Base Leg

While the aircraft is on the base leg, autoland uses a similar updating strategy to the method
on the downwind leg. The updating strategy is shown in Figure 2-15. Since the final leg
cannot be moved parallel to its current location, as was done with the base leg for trajectory

57
 p c1 d
Finally,
α = χautoland
χ adjusts d0requirements
3 −Functional for the actual L/Dav .
Figure 2-16: Example of trajectory updating3. on thethe
Hold base leg
aircraft after
on afirst update
holding pattern by the initial �
FB,north
�point=on d cos(χ
until
1. Select destination -Engine
FB,north dthe
=guid c1
co
within a bandThe energy1.required
of autoland Select isdestination
function designed
to to allowthe
commence an aircraft
standard
-Use
to semi-auto
Holdingtraffic pa p
2 traffic
2. Guide the aircraft to an initial point on the standard Design
� pattern atConcept
thethrottle
selectedcommands) perform an emergency landing
destination � L/D
fromav any point
Finally, autoland adjusts d0 for the actual L/Dav FB,north
. = d cos(χc1 ) + F2.B,north
autoland needs to employ several 1 d =
Guide the aircraft to an initial
functions. Equations
d 0F · B,east = d sin(χ
point on th
� 10
c1

3. Hold the aircraft on a holding pattern by the initial point on

-Engine until the aircraft’s
guidance energy
to1.Initial has reduced
Point -Engineto off approach FB,east = d si
from Initial
within a band of energy required to commence the-Usestandard traffic
Holding pattern
pattern -Use
Select destination
3. Hold the aircraft on aF6F,north
trajectory updating for
3 management
�holding
energy
=Functio
dpattern
cos(45◦ )by
co
L/DF �av F=FIdissin(χmoved2. ) moved
+ Fthe
c1Guide
within
outwards
aircraft
B,east
atoband of energy
analong
initial point onrequired
course χ
the if dtois comm
c2 ,standard posit
traffic
d = d0 · B,east (2.19) � The autoland◦ fu
10 χ Guide on FF,north == d pattern
cos(45 �
1 1Equations of χc2 , if d is1
Equations 6 3 � Functional
4. caircraft
3. Hold the
Equations
negative. αAfter
theathe F by the
holding pattern
aircraft
first along
d cos(45
initial
the Ftraffic
update,
throttleFthecomman
remains
) sin
point F,east

◦ requirements pof(1)
within a band energy required
energy deviates fromtothe
commence
expected standard
energy
1 outwards
Equations F = d cos(45 ) cos(χ 1 )Equations
+F autoland needs t
αF,north
p (2)
c1 the aircraft χd =updating
atan2(∆t �
F FI is moved moved along course χF,north d threshold.
c2 , if d is positive, and 4.
= f (∆E/W, χinfGuide
ithe
L, αopposite
p , L/D
5. Perform
direction
along the traffic pattern
av )athe
F_FI’ flare maneuver
� while
FtoF,east
as the=aircraft
d cos(45 re
500 ft F_F=F_F’ 500 ft
The autoland function
d 0 =is fdesigned
energy
χd deviates
(∆E/W, d to
αL,allow
from
α p ) anexpected
aircraft energy semi-autonomo
d = f (∆E/W,
of χc2 , if d is F_FI,0 After the first update, Fdcommands)
= f (∆E/W, L, αat
p , L/D ) from 1. point
Select
(1) dest
χcnegative. F
throttle
�
remains
FF,east needs
fixed
= d cos(45
perform 500αavan
◦ 5. Perform
) sin(χ
ft a
◦ the runway
p =emergency
15flare landing
maneuver
6.+Capability
)= as the from
aircraft any
reaches
FF,east to absorb errors in wind measu the of
flare ini
alt
χc autoland Useto of
employ
lookup χcαc1
several
ptables
◦
45functions.
for trajectory updating
threshold. χ αpd(1) F_F=F_F’ F_FI,0 � on
2. Guide the
the
fi α
6. Capability to αpαL/D
◦ absorb errors in wind measurements
= 75 av χd =estima
and unce atan
L
αp (2)αp α 10 d = d0 p7. Capability
L to absorb errors in L/D 0
F_FI’ 1. Select � destination 7. Capability to absorb errors in L/D0 estimation 3. the
Hold the a
α χ χd Since χc2explicit
no
= atan2(∆E, ∆N
χ expression
) can be found up tofor amount
b L 8. 2 b8. Capability
Design
Capability to absorb
to absorb
Concept of 30% error
within
up to of 30% error in an initial (L/D)
in a
a exp
b
Use of lookup tables for trajectory updating
Lack of energy: Excess on the
energy:
2. Guide χc1 tolegan
the aircraft
downwind leg
α and
leg
initial
(1) base
and
pointbase leg
on leg sifto
the itstandard
takes 10 straffic
ato accumula
patte
α (1)
χc2 2 pDesign Concept -> d positive moved to correct
and a given
base leg
2◦ Design Concept
p energy
if it takes 10error ∆E/W
accumulate at any
stable, newloc
L
-> d negative LαDp(L/D)
= -Engine
15 +onL
This
Thisdnumber (L/D)
number
guidance
Bwas to was
from derived
b Initial
derived Point
the from
-Engine
range the
of L/D range o
offmeasurem
approac
χc1 αp (2) for3. the
L/D
Hold the
for
=
av aircraft
any on
αupdates a◦α
-Use
simplythat
(2) Vpattern
pusing
holding
Holding Tpattern
simply
have as /V
using-Use
.V by
T asthe
trajectory
Applying
previously initial point
. updating
wind
/Voccurred,Applying for
corrections until
energy
should
wind
autoland uim
co
Since no explicit expression can beto found amount p =
d that 45
theL leg+offixes
L should tobe
sink sink
-Engine
◦α on guidance Initial Point -Engine offband
approach from
F_B -Engine
get Dα
Initial
rid on guidance
B high
Point
the -Use Initial
standard
frequency Point
traffic
spikes. Hence -Engine
pattern
+/-30% off appr
isHence
consid
αp =F_B15 within a of energy
αp management
= 75 2fortrajectory
◦ required
get
-Use Holding Design
rid to
of commence
the high Concept
the
frequency
pattern -Use trajectory
standard
spikes. traffi
updating for ene
-Use Holding pattern -Use trajectory updating forapproximate
energy
moved to α = 45a◦ given energy error ∆E/W at any tolocation
correct p
along
9. 3 dFunctional
the
Capability togiven
absorb ∆E/W
15% error, inL,
and
requirements αp expected
(L/D) and dafter
t otal if
having
◦
αp L c −9.
= target
χ(L/D) χ Capability to This
+ L point).
touchdown (L/D) absorb 15%
choice is aerror in (L/D)
compromise betwe
ex
αp = 75
for any updates that have previously occurred, autolandchosen
L/D uses over Dusing
=
av pre-populated andtarget
-Engine
the target lookup
B
ontables
optimizer to
touchdown
touchdown point
b
guidance
solve
andpoint). to
the
tolerance Initial
implicit
This
toan L/D
Point
choice
errors equat
in av
is(L/D
a cs
3 Functional requirements 3
The Functional
autoland
L +
function
L
is requirements
designed to allow aircraft to
Figure 2-15: Trajectory
to approximate d for givenupdating
∆E/W , L, method on table
αp and dt otal the 2 Design
measurement,
base autoland
providesleg
if applicable. a This
D
throttle-Use
theConcept
uncertainty
commands)
moreneeds
B
Holding
target
reliable
approach
in
perform
and
was
wind
pattern
touchdown 5point
and errorand
-Use
an emergency
faster functions.
in L/D
trajectory
landing
solution than an o
measua
tolerance
0 from

2 Design Concept to employ several

measurement,
The autoland
The autoland function is designed to allow an aircraft to semi-autonomously function
(with uncertainty
the ispilot
designed to in
following wind
allow an and
simple erro
aircraft
chosen 2 Design
over using Concept
an optimizer
throttle toperform
commands) solve the great
implicit landing
an emergency equation
from anyα
-Engine
on the =
pon
importancefly,
point c1 −
χofSelect
1.guidance
when
throttlesince χitdthe
commands)
initiation to Initial
comes perform
atlookup
destinationaltitude. Point
to theanfinal
For -Engine
emergency
this off fro
landing
purpose, ap
implementation
-Engine on guidance
autoland to Initial
needs to employ
than anprocessor.
several Point
2.
functions.
4Guide
Effect
Furthermore, any
thewhich
-Use
-Engine
of off Holding
approach
istooferror
level
aircraft of
an
3from
autoland
pattern
accuracy
initial
Functional
needs
in can
point
to employ
-Use
Initial
windbe
on the
several
trajectory
Point
me
achie
standar
-Use requirem
functions.
updating
standardfortra
e
updating on the downwind leg,
table provides
-Engine autoland
a on
-Use
more
Holding needs
reliable
guidance to
pattern to insert
andInitial
faster
-Use
solution
Point another
-Engine
trajectory off� leg
updating forinto
optimization
approach the
from
energy traffic
function,
Initial
management pattern
Point -Use standard 5
traffic
�
FB,no
great importance
1. pattern
-Use Holding
when
Select destination
it comes -Use trajectory
to the updatingFfor
final implementation
= dmanagement
ofenergy
B,north cos(χ
3.1.Hold
the system The
)+
Select
c1the FB,north
destination
onprofile aircraft
anautoland
embeddedwith altitude
onfunction
a holding pattern byon
is designedthe tou
initia
to
within a band of energy required 54
the 3 be Functional requirements to commence the
trajectory in between the base leg 2.and Guidethe finalto leg,
the aircraft which
an initial point onwillstandardtermed final
traffic 2.pattern
Guide atthe
throttle
Whenthe
intercept
aircraft to destination
thecommands)
autoland selected an initial point an
perform on the stan
emerg
plansif the length of the legs for� t
processor. Furthermore,
3 Functional any level of requirements
accuracy can be achieved � by reducing autolandstep needs
size, to employ several Ffunc
FB,east = d sin(χ
at
3. the ) the
+ Faircraft
c1aircraft’s current
B,east B,e
on a altitude to predict the in
g
3 creates
leg. The final intercept leg Functional fixrequirements
3. Hold the aircraft
awithin
newa band of
on a holding pattern by the initial point until the aircraft’s energy has reduced to
referred to toascommence
energy required F FI The which
the
autoland
standardwill
the be
functiona located
aircraft
within
traffic is is
band
pattern ondesignedatrequired
theenergy
of to allow
downwind
Hold
anthe
legtoand aircra
baset
commence
holding pattern by the

The autoland function is designed 54 to allow an throttle

aircraftcommands)
tothesemi-autonomouslyperform
1. Select
trajectory an emergency
destination
to account (with
for anythe landing
pilot
changes � in
5 f
follo
win
The autoland function is designed to allow Fan� aircraft to semi-autonomously
= d cos(45 ◦ Guide
4. ) cos(χ the ) + F
aircraft (with
along the
the pilot
traffic Ffollowing
pattern whil=
the intersection between the throttle
base leg 4. and theperform
Guide
commands) the aircraft new final
along the traffic
an emergencyintercept
pattern
F,north while
autoland
landing leg.from
updating The the
needsany final
traffic
to
energy
employ
point c1 leg
pattern
of
deviates
fix
length
several
initiation
F,north
fromat
is
as the
at
thealtitude.
aircraft’s
functions.
altitude.
expected
F,north
energy
For th
throttle commands) perform
energy deviates from an
autoland needs to employ several functions. emergency
the expected energy landing from any point of
2. initiation
Guide the aircraft to For
an this
initial p p
moved to a point 500 ft out autoland
fromneedsthe torunway
5. Perform employ several
threshold,
a flare maneuver functions. which
as the aircraft �reaches is 1. the
theSelect
flare latest
destination
5.
◦ Perform
altitude position,
above a the
flarelanding the
maneuver site as the aircraft F � reaches
=
FF,east = d cos(45 ) sin(χ 3. Holdc1 ) + F the aircraft on a holding
F,east F,east
1. Select destination 6. Capability to absorb errors in wind measuremen
6. Capability to absorb errors in wind measurements and uncertainty in within
wind a
profile band
with of
altitudeenergy require
final leg may be intercepted 1.after
Select destination
an update on the base leg. 500 2. ft Guide
was chosen, the aircraft since to anthis initial point on the st
2. Guide the aircraft
7. Capability to absorb toerrors
an initial
in L/D0point on the
estimation χ�d = standard traffic
7. Capability
atan2(∆E, ∆N pattern
to) absorb at theinselected
errors L/D0 estimation destinχ
2. Guide the aircraft to an initial point on the standard 3. Hold traffic the 4.pattern
Guide onat the
a the aircraft
selected along the
destination tra
◦ aircraft holding pattern by the
is equivalent to the distance required 8.
3. Hold to intercept
χdCapability
the to absorb
aircraft onupathetoholdingfinal
of 30% error leg
in an from
pattern initial
by the a initial
(L/D)
within 45 8.
expected
angle,
χadCapability
whileenergy
band
point the
of towhich
absorb
aircraft
energy
until the up
deviates istotheoffrom
isrequired
on
aircraft’s 30%toerror
downwind the inexpect
commence
energy anhas
init
3. Holdwithin
the aircraft
αpleg= and ◦base on
a15band ofifaenergy
leg itholding
takes 10 spattern toby
to accumulate
required the initial
a stable,
commence newthepoint
L/D leg
=anduntil
15◦base
p estimate
αstandard the
(drives
traffic aircraft’s
leg if it takes
choice
pattern legenergy
of 10 s lengths).
to accumulatehas redu as
the maximum course differencewithin thatαacan This
band occur
number with
was
◦of energy derived the
requiredfromtrajectory
the
to range
commenceof L/D
= 45using VT as /Vsink . Applying wind corrections should improveupdating
measurements
the standard
α p
strategy
This
= 5.number
taken
simply
◦Perform
traffic
45 during on
was
pattern the
anderived flarefrom
aAvidyne flightthetest,
maneuver rangeas of the
L/D
psimply theusingaccuracyVT as /V
and sink . Applying
averaging wind correcti
should
4. Guide
αpget the
= rid
75of◦ aircraft along the traffic pattern while updating αpget =to6.75be◦of
rid the high traffic pattern length
Henceas th
the high frequency spikes. Hence +/-30% is considered athe
Capability
pessimistic frequency
to absorb
estimate. spikes. errors +/-3
in w
base leg. This point will become the energy new final from
deviates leg fix, F F and
the expected will remain at this position
energy
9. Capability to absorb 15% error in (L/D)expected after 9. Capability to absorb 15% error in (L/D)
5 having intercepted
target7. touchdownthe final leg (drives choice of
Capability point).to absorb errors
expected
in L
during all subsequent updates to5.2the trajectory,
Perform
target atouchdown
Design while
flare maneuver
point). This
Concept on theaircraft
aschoice
the isbase leg.
a compromisereaches 2
Only the Design
between fix
flare
runway F
the target
FI
altitude
consumed Concept
will
touchdown
above
due beto This
the choice
placementlanding ofis a site
compr
the target touchdown point and tolerance to errors in (L/D)expected originating frompoint errors and in tolerance
wind to er
8. Capability
measurement, uncertaintyto absorbin wind up
andto of 30%
error in
6. Capability
measurement, to absorb
uncertaintyerrors in wind
in wind and errormeasurements
in L/D◦ 0 measurement. and uncertainty intowind profile with altL
moved inwards or outwards along course
-Engine onχguidance
c2 , which to is angled
Initial Pointat 45
-Engine tooffthe-Engine final
approach onfromleg
guidance
leg course.
and base
Initial Initial
leg
Point if it
-UsePoint
takes -Engi
10
standard st
7.-Use Holdingtopattern
Capability absorb-Use errors trajectory
in L/D0updatingestimation -Use
for Holding
energy This
managementpatternnumber -Usewas trajectory
derived updat from
As with the updating strategy on 4 the Effect downwindof errorleg, in the wind amountmeasurement d that4 the Effect
and final of
uncertainty error inofwind
intercept
simply using VT as /V wind measu
sink . Apply
profiletowith
8. Capability absorbaltitudeup to of 30% on error
touchdown in an initial point profile
(L/D) getwith
expected of altitude
rid while thethe high on touchd
aircraft
frequency is on sp th
fix F FI should be moved, depends3leg onand Functional
thebase distanceleg if it to requirements
go 10
takes until
s to the fix, L,athe
accumulate 3stable,
When
Functional
angle new
autoland
αL/Dp thatestimatethe requirement
(drives choice of l
When autoland plans the length of the legs for the traffic pattern trajectory, it plans
uses the
9. Capability
thewindlength of the legs for the tra
measurements
toduring
absorb 15% error
This number was derived from the range of L/D
at the aircraft’s current altitude to predict the glide ratio that
measurements
atthe
theaircraft
aircraft’s can current
achieve
taken
altitude
on each to an
predict
leg. While
Avidyne
the glide r
aircraft course makes with courseThe χsimply and The autoland function theis accuracy
designed tobase allow a
the autoland
c2aircraft isthe
using VTaverage
on function
the /Vsink
downwind
as glide
is designed
.leg
Applying
and theratio
to wind
base legexpected
allow an
theaircraft
ofcorrections
traffic on to the
should
thepattern,
aircraft remainder
istarget
semi-autonomously
improve
on
autoland touchdown
the can
downwind
adjust leg(with
the point).
and
length of This
thethe
and leg ofcf
pilot
avera
throttleridcommands)
gettrajectory
the of the high perform
to account frequency anspikes.
for any changes emergency
in wind andlanding
Hence +/-30%
hence
throttle
thefrom
changes
commands)
any
trajectory
isinconsideredthe
aircraftpoint target
to account
glideofto perform
touchdown
initiation
for
ratio.be any
As
an
changes
a pessimistic
soon
emergency
point
atasaltitude.
in wind
the and
and
estim F
of the trajectory L/Dav . In addition autoland to needs
that,todemploy depends severalonfunctions.
the distance autoland
dtotal needs between to employ
measurement, several functions
the uncertainty in wi
9. Capability to absorb 15% error in (L/D)expected after having intercepted the final leg (driv5
5
initial location of the final intercept target
1.leg touchdown
fix,destination
Select F FI,0, point).andThis choice location
current is a compromise of1.FSelect between
FI destination
(i.e. runway
beforeconsumed due to p
the target touchdown point and tolerance to errors in (L/D)expected originating from err
2. Guide the uncertainty 2. Guidetraffic the aircraft toatanthe initial point
the update). This is due to the fact that the aircraft
measurement, angle between to an in initial
windχand point
c2 and erroron the
the
in L/D standard
course of thepattern
0 measurement. final selected de
3. Hold the aircraft on a holding pattern by the3.initial Hold point the aircraft until the on aircraft’s
a holding energy patter
intercept leg, χf i does not remain constant over the region of trajectory
within a band of energy required to commence the standard traffic pattern updateswithin aas bandfor ofthe energy required to c

downwind leg but changes with change in dtotal . In Figure 2-15 dtotal is equal to zero. Figure
2-16 shows a non-zero dtotal for the case that the trajectory had previously
5 been elongated
5
(blue curve) and the next update intends to shorten the trajectory (red curve).

58
 αp (2)
α

LD (L/D)d + LB
L/Dav =
LD + LB

αp = χc1 − χd
The distance d by which F FI needs to be moved to remove energy error ∆E/W is

1 Equations
�
FB,north = d cos(χc1 ) + F
where γ = π− | (χ1 − χ2 ) |.
1 Equations
�
−−−−−−→ χc
FB,east
F_FI’ = d sin(χ c1 ) + F
d = f (∆E/W,
dtotal =| FF I,0 , FF I | (A.58)
L_FI χc
1 d Equations
αp
α �
d_total
αp (1) FF,north = d cos(45◦ ) cos(χdc1
0
αp (2)
F_FI’
α L
F_F=F_F’
2 Design
�
FF,east =Concept
d cos(45◦ ) sin(χ
L (L/D
c1 D
L/Dav =
F_FI,0 χb
L_F=1,500 ft -Engine on guidance to Initial Point -
αp (1)
Previous energy corrections
α-Use
p (2) χ�d =-Use
Holding pattern trajectory
atan2(∆E, αp =
∆Nu
χ
Aircraft
created d_total >0 α
position, P
χc2
Next update for lack of energy: 2 Design Concept
-> d negative χc1 3 Functional requirem L/Dav =
◦ -Engine on guidance to Initial Point -Engine off appr
αp = 15
Thepattern
-Use Holding autoland
-Usefunction
trajectoryisupdating
designedfortoene
a
αp = 45◦
throttle commands) perform an emerg
αp = 75◦
autoland needs
3 Functional to employ several func
requirements
Figure 2-16: Example of trajectory updating on the base leg after first update �
FB,north
The autoland1.function
Select isdestination
designed to allow an aircraft
2 Design Concept
throttle commands) perform an emergency landing fro
F �
autoland needs to employ
2. Guide theseveral functions.
aircraft to an B,east
initial p
calculated in the same manner as for the-Engine
downwind leg using look-up tables.
on guidance to1.Initial
The only
Point -Engine off approach fro
Select destination �
-Use Holding pattern -Use 3. Hold updating
trajectory the aircraftforonenergy
aFF,north
holding= dp
man
difference is the added dimension of the look-up table due to the dependence on d to an
2. Guide the aircrafttotal
. initial point on the stan
within a band of energy required
�
F by the
3. Hold the aircraft on a holding pattern = in
d
4. Guide the aircraft alongF,eastthe traffi
3 L,Functional
d0 = f (∆E/W, αp , dtotal ) requirements
within a band of energy required to commence t
(2.16)from the expecte
energy deviates �
χd
4. Guide the aircraft along the traffic pattern whil
The autoland function χd is designed
energy to allowa flare
deviates
5. Performfrom an aircraft
the expected to as
energy
maneuver semi-
the
with throttle commands) perform αan
p =emergency
5. Perform
◦
landing
a15flare maneuver from any
as the aircraft p
reaches
α 6.
= Capability
45 ◦ to absorb errors in w
autoland needs to employ several functions.
p
6. Capability
αp = χc2 − χb αp = 75to ◦ absorb errors in wind measuremen
7. Capability(2.17)
to absorb errors in L/
1. Select destination 7. Capability to absorb errors in L/D0 estimation
Then, 8. 2 Design
8. Capability
Capability Concept
to absorb upto
to absorb up to
of 30% error in of
an 30%
init
2. Guide the aircraft tolegan
andinitial
legleg
base point
andif itbaseonleg
takes the
10 itstandard
sifto takes 10 stra
accumulate ato
s
L/Dav -Engine onThis
This number guidance
wasnumber to Initial
derived Point
from derived
was the -Engine
range of L/D
from t
d = d0 ·3. Hold the aircraft on-Use
a holding (2.18)
10 Holding
simply Tpattern
using Vpattern
as /Vsink-Useby the
trajectory
. Applying initial
wind po
updating
correcti
simply using VT as /Vsink . Applyi
get ridrequired
within a band of energy of the highto frequency spikes. the Hence +/-3
get rid of commence
the high frequency stand
spi
with 9. 3 Functional
Capability to absorb 15%requirements
error in (L/D)expected
L(L/D)b + LF I (L/D)f i target 9. Capability
touchdown to This
point). absorb 15%
choice is aerror
compri
L/Dav = the target
The autolandtouchdown
function
target (2.19)
point andpoint).
is designed
touchdown tolerance
to allowtoan
This er
ch
L + LF I measurement,
throttle uncertainty
commands) performin wind 5 error in
and
an emergency lanL
the target touchdown point and
autoland needs to employ several functions.
The length of the final intercept leg LF I measurement,
depends only on the current location uncertainty in win
of the final
188 1. Select destination
intercept leg fix, F FI. The location of the final leg fix F F is fixed at 500 ft from the
4 Effect of error in w5
2. Guide the aircraft to an initial point on
runway threshold. Hence, the total length of the final leg LF is 1,5003. ftHold
after adding the
the aircraft on a holding profile with altitude pattern b
within a band of energy required to com
59 When autoland plans the length of the
at the aircraft’s current altitude to pr
the aircraft is on the downwind leg an
the trajectory to account for any chan
displacement of the target touchdown point from the runway threshold as discussed in more
detail in chapter 2.1.11.

Use of Lookup Tables for Trajectory Updating on the Downwind Leg and Base
Leg

Since no explicit expression can be written for the amount d that the leg fixes should be
moved to correct a given energy error ∆E/W at any location along the trajectory and for
any updates that have previously occurred, autoland uses pre-populated lookup tables to
approximate d for given ∆E/W , L, αp and dtotal , if applicable. This approach was chosen
over using any optimization function to solve the implicit equation on the fly, since the
lookup table provides a more reliable and faster solution than an optimization function,
which is of great importance when it comes to the final implementation of the system on an
embedded processor. Furthermore, any level of accuracy can be achieved with the lookup
table by reducing the step size, assuming memory is not restricted. Since the slopes of the
functions d0 = f (∆E/W, L, αp ) and d0 = f (∆E/W, L, αp , dtotal ) change slowly, however,
memory requirements for the lookup tables are expected to be limited.
Figure 2-17 shows a conceptual view of the implementation of a 3D lookup table for
d = f (∆E/W, L, αp ). The values of d for each combination of entries from vectorized ∆E/W ,
L and αp are found by sweeping across the entries of one vector, while keeping the other two
values constant. For constant αp (1) and L(1) for example, the values of d for a vector of
∆E/W can be stored in a one-dimensional array. Continuing this process creates a three-
dimensional array, whose entries can be imagined to be pinned to a point in 3D space. For
equal spacing of the intervals in all three dimensions, the arrangement can be visualized
by a square as shown in Figure 2-17. The point in space is defined by the x-, y-, and z-
coordinates representing ∆E/W , L and αp and the value at this point is d. In order to
find d for a combination of values of ∆E/W , L and αp that falls in between the grid of
stored values, autoland uses a simple linear interpolator to interpolate the function in 3D.
A linear interpolator is considered sufficient due to the slow change in slope of function

60
1 Equations
d0 = f (∆E/W, L, αp , dtotal ) (1)

d = f (∆E/W, L, αp , LF , (L/D)av ) (2)

1 Equations
d = fd(∆E/W, L, αpL,
= f (∆E/W, ) αp , L/Dav ) (3) (1)
χc
L/Dav
αp (1) d = d0 (4)
1
αp (2)
Equations 10
χf !E/W(1) !E/W(2)
◦
α d = f (∆E/W, L, αp , L/Dav ) (1)
αp = 30 L(1) d11 d12
αp (2) χ
L(2)
c d21 d22
LD (L/D)d + LB (L/D)b
L/Dav = (2)
α αp (1) LD + LB
αp (2)
LB (L/D)b + LF I (L/D) + χLF (L/D)f
αp = χfci − (3)
α L/Dav = !E/W(2)
!E/W(1) (5)
LB + LF I + LF
L(1) d11 d12 LD (L/D)d + LB (L/D)b
2 Design Concept L(L/D)dL/D + Lav =
B (L/D) b + LF (L/D)f (2)
L(2) d21 L/Dav =d22 LD + LB (6)
LD + LB + LF
-Engine on guidance to. Initial Point -Engine off approachαp = χc from
− χ Initial Point -Use standard traffic pattern (3)
-Use Holding pattern -Use. trajectoryLupdating
F I (L/D)ffori +energy
LF (L/D)
management
f
. L/Dav = (7)
LF I + LF
2 Design Concept
3 Functional requirements
αp = χf − χf i (8)
Figure 2-17:
-Engine ConcepttoofInitial
on guidance a 3DPoint
lookup table
-Engine off for determination
approach from Initial of d on
Point downwind
-Use leg
standard traffic pattern
The-Use
autoland function
Holding is designed
pattern to allowupdating
-Use trajectory an aircraft
forto semi-autonomously
energy management (with the pilot following simple
throttle commands) perform an emergency χc2 − χbfrom any point of initiation at altitude. For this
αp = landing (9)purpose,
d= autoland needs
f (∆E/W, L, αp ).to employ several functions.
3 Functional requirements αp = 110◦ (10)
1. Select destination
The autoland function is designed to allow an aircraft to semi-autonomously (with the pilot following simple
2.throttle
Guide commands)
the aircraft perform
to an initial GP Spoint err on the 2m standard traffic pattern at the selected destination
∆d = an emergency = landing ◦)
from
= 300fanyt point of initiation at altitude. For
(11)this purpose,
Final 3.Leg
autoland needs to employ tan
several γ ffunctions.
lare tan(1.5
Hold the aircraft on a holding pattern by the initial point until the aircraft’s energy has reduced to
within a band of energy required � to commence the standard traffic pattern
1. Select destination FB,north = d cos(χc1 ) + FB,north (12)
4.
Up until theGuide the
aircraftaircraft along
beginstothe the traffic pattern while updating the traffic pattern length as the aircraft’s
2. Guide the aircraft an turn
initial onto
point the
on the final leg, the
standard traffic trajectory
pattern at can be updated
the selected using
destination
energy deviates from theFexpected �
= denergy
sin(χ ) + F (13)
B,east c1 B,east
the strategies
3. Holdjustthepresented.
aircraft on aAs soonpattern
holding as thebyaircraft begins
the initial
5. Perform a flare maneuver as the aircraft reaches the flare altitude above the landing site
point to turn
until the onto theenergy
aircraft’s final leg, how- to
has reduced
within a band of �energy required to ◦
commence the standard traffic pattern
ever, there is no means Fto shorten
F,north = d cos(45 ) cos(χc1 ) + F
themeasurements
trajectory anymore,
F,north even though elongation (14) of the
6. Capability to absorb errors in wind and uncertainty in wind profile with altitude
4. Guide the aircraft along the traffic pattern while updating the traffic pattern length as the aircraft’s
trajectory would
energystill
7. Capability be possible.
deviates
to absorb from
Ferrors
� theinexpected ◦energy
L/D0 estimation
F,east = d cos(45 ) sin(χc1 ) + FF,east (15)
8. 5. Perform
Capability toa flare
absorb maneuver
up to of
One means to counteract the problem of not being able as
30% the aircraft
error in an reaches
initial the
(L/D) flare altitude above the landing
to shorten the trajectorydownwind
expected while the aircraft is on site
the on the
leg and base leg if it takes F � 10 s to accumulate a stable, new L/D estimate (drives choice of leg lengths).
= d cos(χ ) + F (16)
6. Capability to absorb errors in wind measurements and uncertainty in wind profile with altitude
F I,north c2 F I,north
final leg would
This numberbe towasplan derivedto turnfrom the onto theoffinal
range leg with excess
L/D measurements takenenergy
during anand depleting
Avidyne the
flight test,
simply using VTto
7. Capability as /V sink . errors
absorb Applying in wind0 corrections
L/D estimation should improve the accuracy and averaging should
excess energy
get ridthrough
of the high S-turns FF� on
frequency I,east the = dfinal
spikes. sin(χ
Hence leg +by FFdefault
c2 )+/-30% if the aircraft
is considered
I,east performed
to be a pessimistic nominally.
(17)
estimate.
8. Capability to absorb up to of 30% error in an initial (L/D)expected while the aircraft is on the downwind
9. Capability
A straight final legto
leg and absorb
would
base leg if15%
then
it takes
�
error 10ins be
only (L/D)
flown
to accumulate
expected after
if the having
glide
a stable, new intercepted
performance
L/D estimate theis
final
equal
(driveslegchoice
(drives
of choice
to (18)
the worst
leg of
lengths).
target touchdown point). FF,north
This = LI cos(χ
choice is a f ) + FF,northbetween runway consumed due to placement
compromise of
This number was derived from the range of L/D measurements taken during an Avidyne flight test,
glide performance
the simply designed
target using
touchdownVT as /V for.
point However,
and
sink . Applying
tolerance
wind S-turnserrorsor
tocorrections in other
(L/D) trajectory elongation
improveoriginating
shouldexpected the accuracyfrom strategies
anderrors in wind
averaging should
measurement,
get rid of the uncertainty
high frequency in windspikes.and error
Hence in +/-30%
L/D0 measurement.
is considered to be a pessimistic estimate.
require a non-significant amount of length
5
of the straight line final leg to perform, since
9. Capability to absorb 15% error in (L/D)expected after having intercepted the final leg (drives choice of
the aircraft target
will need to bepoint).
touchdown realigned withisthe
This choice course ofbetween
a compromise the final leg.consumed
runway Assuming
due tothat the of
placement
the target touchdown point and tolerance to errors in (L/D)expected originating from errors in wind
turn radius is fixed, as would
measurement, be the
uncertainty caseand
in wind when
errorinterfacing to the existing autopilot without
in L/D0 measurement.
5
modification, a 5% elongation of the trajectory would require two turn radii of distance to

61
5
complete as shown in Figure 2-18. This is equivalent to correcting a 5% error in energy on

Figure 2-18: Percent change in length of trajectory and amount of final leg length required
for range of angles turned away from trajectory

the final leg.

With a nominal turn radius of 1,100 ft and a final leg length of 6,000 ft from the final leg
intercept fix F FI,0 to the target touchdown point, there is a maximum of three turn radii of
final leg length available for S-turn maneuvers after subtracting the turn from the base leg
to the final leg of 1,100 ft, the intercept margin of 500 ft and the margin for the touchdown
point dispersion of 1,000 ft as shown in Figure 2-6. With three turn radii, a maximum of

62
15% of error in energy could be corrected, which would be sufficient for the energy errors
expected. However, this energy error would need to be known by the start of the final leg.
As the shape of the curve in Figure 2-18 suggests, the capability to correct energy errors
decreases rapidly with decreasing distance available to perform the maneuver. Since it is
assumed, that significant energy errors will not be visible at the start of the final leg but
appear later, the S-turn maneuver is expected to have insufficient capability to correct en-
ergy errors and hence will not be included in the current design. However, it is a means to
improve performance in a second iteration of the design, if deemed necessary.
The approach chosen instead, is to place the target touchdown point in a way that ex-
pected energy errors can be carried through until touchdown, without the aircraft landing
short of the runway for the maximum negative energy error expected, or running off the end
of the runway for the maximum positive energy error expected.

2.1.10 Placement of the Target Touchdown Point

The energy error correction strategies on the downwind leg and base leg aim to remove any
energy error before the start of the turn onto the final leg. The assumption is therefore, that
only energy errors accumulated during descent along the final leg will need to be absorbed
by placement of the target touchdown point. Any energy error on the final leg is due to
an error in estimating the expected glide ratio of the aircraft on the final leg, L/Df . This
error in estimation may either be due to an error in measuring the aircraft’s current glide
performance, L/D0 , or due to an error in predicting wind shear during the descent along the
final leg. Both error sources will lead to an error in expected glide ratio, L/Derr . This error
will cause overshoot of the target touchdown point if the glide ratio was overestimated and
will cause undershoot if the glide ratio was underestimated by the amount ∆d = LF ·L/Derr ,
where LF is the length of the final leg.

63
Overshoot or Undershoot due to Wind Shear in the Boundary Layer

At the start of the final leg, the aircraft is on the order of 500 ft - 800 ft off the ground
depending on glide performance, which is within the region of the atmospheric boundary
layer. Hence, the wind speed is expected to change from the measured wind speed at the
start of the final leg until touchdown. While on the downwind leg and the base leg, the
boundary layer effect would also be present. However, its effect is expected to be less
pronounced. Since errors in glide performance estimation can be corrected on the downwind
leg and the base leg, the boundary layer effect on the wind speed is ignored until the start
of the final leg.
Figure 2-19 shows a sample wind profile derived from the Canadian Global Environmental
Multiscale (GEM) model, which will be considered typical [23]. The change in wind direction
will be ignored for simplicity and only the change in wind speed considered. The figure shows
that the wind speed changes from 12 m/s (25 kt) at 300 m (1,000 ft) to zero on the ground.
The strongest change in wind speed occurs below 600 ft, which suggests the importance of
including its effect for glide performance prediction on the final leg. Above 600 ft, the effect
is smaller and it is expected that resulting energy errors can be corrected via trajectory
updating. In order to simplify the calculations, the wind profile in the boundary layer will
be approximated by a linear model as shown in Figure 2-20. The wind profile is fixed by
parameters h0 and Vw , which indicate the start height of the wind shear and the wind speed
at this height respectively. Using this model, an expected glide ratio can be calculated,
assuming the aircraft descends through the simplified linear profile. The average glide ratio
experienced under these conditions will be denoted by L/Dshear , which can be expressed
in terms of the expected glide ratio experienced without the boundary layer effect present,
L/Df , the length of the final leg, LF and the amount of overshoot due to the change in wind
speed, ∆d. ∆d is positive, i.e. is an overshoot, for a reduction in headwind. ∆d is negative,
i.e. is an undershoot, for a reduction in tailwind.

∆d
L/Dshear = L/Df (1 + ) (2.20)
LF

64
 288 JOHN D. WILSON AND THOMAS K. FLESCH

Figure 2-19: Example of change in wind speed, S and direction, dir with altitude in the
atmospheric boundary layer

altitude above ground

h_0

Figure 1. Fitted analytical wind profile (lines) versus output profile from the GEM numerical weather
model (symbols). The lower panels are a ‘blow-up’ for the surface layer. Orientation of the coordinate
system ensures V = 0 in the surface layer.
Figure 10: Change in wind speed and direction with altitude as per GEM model da

Figure 10 showshs = 100the

that δ = 500 in
m, change L = 5000
m,wind speedm)with
yielding "% S can
altitude = 5.6% and "be
roughly β
4◦ ; in
=approximate
general0 the
function from altitude uptwo-layer model
until about 300gave
m excellent
above thefits ground
to simpler profileisshapes
which on thelikeorder
this of th
wind speed
one (either a monotonic increase with height in windspeed, or a low-level
which the downwind leg would be started. Figure 11 illustrates the linear function of wind speed
V_w,0 jet giving
that will be used aforsingle local speed
analysis. Sincemaximum).
a direct headwind or tailwind provides the worst case un
The second example
point(b) on Figure 2 shows a slightly more complex profile
Figureovershoot of the
2-20: Linear target touchdown
approximation of change inrespectively,
wind speed for analysis
in the purposes
boundary layerit will be assumed t
shape (both a local maximum and minimum S), and an intermediate quality of fit
direction is aligned with the aircraft heading −1 at at 0◦ or 180◦ . %
(for which u∗ = 0.06 m s , hs = 5 m, δ = 100 m, L = 10 m) with " S =
17% and "β = 9.6◦ . Example (c) is even more complicated, the GEM windspeed
The magnitude of ∆d can be calculated as follows. The wind speed Vw at any altitude h
above the ground can be calculated from knowledge of the wind speed Vw,0 at altitude h0 at

65
the start of the final leg.
Vw,0
Vw (h) = ·h (2.21)
h0
The change in distance travelled over ground due to this wind shear can be found from:

t(h=0)
�
∆d = Vw (t)dt (2.22)
0

Changing the integration variable to h and plugging in the equation for Vw (h) yields:

�h0
1 Vw dh
∆d = h · dh dt = (2.23)
Va sin γ h0 Va sin γ
0

Integrating and considering that:

LF
sin γ = (L/D)−1 L/Df = (2.24)
h0

results in
1 Vw,0
∆d = · LF · (2.25)
2 Va
Equation 2.25 shows that the magnitude of overshoot or undershoot due to a linear wind
profile is linearly proportional to the length of the final leg, linearly proportional to the
change in wind speed from h0 to 0, Vw,0 and inversely proportional to the airspeed Va .
Inserting this result into equation 2.20 gives the simple result:

Vw,0
L/Dshear = L/Df · (1 + ) (2.26)
2Va

Autoland will assume that when measuring wind speed Vw,0 at the start of the final leg, that
this wind speed will reduce to zero by the time of touchdown and assume that the average
glide performance to be expected on the final leg is L/Dshear . In case the wind speed does
not reduce by this amount, the aircraft would undershoot the target touchdown point for an
approach into a headwind or would overshoot the target touchdown point for an approach

66
into a tailwind. The amount of overshoot or undershoot can be calculated from the error in
L/Dexp made by assuming a linear wind profile. From equation 2.26, it can be seen that the
Vw,0
L/Dexp adjustment due to the wind profile is given by 2Va
. The error in L/Dexp is therefore
given by the magnitude of the overestimation of change in wind speed. If the wind was
expected to reduce by magnitude Vw,0 , but only reduces by magnitude ∆Vw , the difference
Vw,0 −∆Vw
Vw,0 − ∆Vw contributes to an error in glide performance estimation of L/Derr = 2Va
.
Figure 2-21 shows the amount of L/D error accrued for a range of ∆Vw if the assumed
reduction in wind speed was Vw,0 . Figure 2-21 assumes the wind speed at the start of the
final leg is 7 m/s. This wind speed would be experienced for a final leg length of 6,000 ft,
glide ratio of 10 and wind speed at 1,000 ft of 12 m/s as shown by the wind model data in
Figure 2-19. If the wind speed then only reduced by 1 m/s, the error in L/Dexp resulting
from making this assumption is 6.8%.

Figure 2-21: L/D error resulting from assuming linear reduction in wind speed of Vw,0 when
wind reduces only by ∆Vw

67
Overshoot or Undershoot of the Target Touchdown Point due to Measurement
Errors

In addition to errors in predicting boundary layer wind shear, errors in estimating the glide
performance on the final leg, L/Df , may originate from an error in the measurement of
L/D0 , or in the measurement of the wind speed and direction. It is assumed that these
measurement errors contribute less than 5% of error to the estimation of L/Df . Figure
2-22 shows the amount of overshoot or undershoot, ∆d that results from overestimating or
underestimating the expected L/Df by the percentage error L/Derr .

1200

1000
Touchdown point dispersion in ft,!d

800

600

400

200

0
0 2 4 6 8 10 12 14 16 18 20
Error in L/D estimate in %

Figure 2-22: Overshoot or undershoot from error in estimating expected L/D

Selected Location of the Target Touchdown Point

For the error in predicting boundary layer wind, a maximum error of 6.5% will be assumed
and for the measurement of current glide performance, a maximum error of 5%. From

68
Figure 2-22, this is equivalent to overshooting or undershooting the target touchdown point
by 400 ft and 300 ft respectively. Since it is assumed that the error in measurement of
glide performance can be both positive or negative, margin for touchdown point dispersion
is required in both directions. For the error in wind prediction, only undershoot margin is
required for the headwind case and only overshoot margin is required for the tailwind case.
Furthermore, bilateral margin for touchdown point dispersion is required for the expected
GPS error of +/- 2 m. Assuming a flight path angle during the flare maneuver of −1.5◦ as
described in chapter 2.1.11, this results in a touchdown point dispersion of +/- 300 ft. Figure
2-23 illustrates the placement of the target touchdown point for the headwind case. For the
tailwind case, the 400 ft allocated to wind shear would be applied to overshoot instead of
undershoot of the target touchdown point.
With this choice, the target touchdown point will be placed 1,000 ft from the runway

L/D = 1.05 L/D_exp

TARGET TOUCHDOWN POINT
L/D_exp=10

600 ft L/D = 0.95 L/D_exp

Undershoot due to
GPS error = + 2 m
L/D = 0.88 L/D_exp Overshoot due to
(-7% from wind shear) GPS error = - 2 m
Latest touchdown
Flare height 400 ft 300 ft 300 ft

6,000 ft

+/- 2 m
!F

Earliest touchdown 300 ft 300 ft

400 ft

300 ft 300 ft

Figure 2-23: Placement of target touchdown point for consideration of L/D estimation error
and GPS error

threshold in the headwind case, in order to ensure that the aircraft touches down at the
runway threshold at the earliest, if all error sources were at their extreme minimum. The

69
latest touchdown would occur 1,600 ft from the runway threshold if all error sources were at
their extreme maximum. For the tailwind case, the target touchdown point is placed 600 ft
from the runway threshold with the latest touchdown occurring also after 1,600 ft.

2.1.11 Flare

Autoland performs a flare maneuver similar to the one implemented by autoland systems
for commercial aircraft, in order to reduce the aircraft’s vertical speed on the final leg to a
vertical speed that is acceptable for touchdown. For an airspeed of 90 kt and an assumed
range of glide ratios from 13:1 to 7:1, the aircraft’s vertical speed ranges from 10 ft/s to 20
ft/s, which is above the 10 ft/s threshold, that is considered to cause damage to the aircraft
and potentially severe injuries to passengers, if a landing was attempted at this vertical
speed [18]. The recommended touchdown speed ranges from 1 ft/s to 6 ft/s for this class
of aircraft. In order to provide a smooth decrease in vertical speed from the start of the
flare until touchdown, autoland uses the following exponential equation, that is shaped by
parameters vtd,0 and τ .
t
v(t) = (v0 − vtd )e− τ + vtd (2.27)

where v0 represents the vertical speed at the start of the flare, vtd the target touchdown
speed at the expected touchdown point, and τ the time constant at which the vertical speed
command is decreased. vtd and τ are chosen in such a way, that the vertical speed at the
earliest and latest expected touchdown point due to GPS error remains within the 1 ft/s to
6 ft/s range and that the airspeed remains above the stall speed even for the latest, expected
touchdown due to maximum negative GPS error, at the worst glide ratio. The onset of a
stall is specified to occur at 73 kt without flaps for the Cirrus SR-22 [8]. For values τ = 1s
and vtd = 3.3 ft/s, satisfactory performance could be achieved. With this choice, the flare
height becomes 70 ft, which is more than twice the typical flare height of 30 ft for this size of
aircraft. However, this is necessary due to the fact that the aircraft is losing altitude twice
as fast as on a standard 3◦ glideslope, which autoland compensates for by starting the flare

70
at a higher altitude.
With a vertical speed of 3.3 ft/s and an airspeed of 75 kt, the flight path angle be-
comes 1.5◦ , which is used to estimate the amount of overshoot and undershoot of the target
touchdown point given the uncertainty in vertical position due to GPS error. Figure 2-24
illustrates the aircraft’s flight path during the flare maneuver for the maximum positive and
maximum negative expected GPS error. For a GPS error of +/- 6.6 ft (2 m), the overshoot

*&+)$,&-%./01)2/&3)%

G#$/0)2"%"&7891&+:%'&$%% 6#")2"%"&7891&+:%'&$%%
CD@% !"#$"%&'%(#$)% ;#<%H>)%?*!%)$$&$% ;#<%=>)%?*!%)$$&$%
A%E@%F"2%
(#$)%9)0.9"%
AB@%'"%
CD"1A%BK%F"2%
45% IJK5%
HL;%
@%
,L;%
4@@%'"%

Figure 2-24: Flare maneuver under uncertainty in vertical position

and undershoot of the target touchdown point becomes +/- 300 ft.

2.2 Autoland - FMS Interface

2.2.1 Trajectory Specification

The trajectory generated by autoland consists of a set of straight line segments for the path
to the initial point and for the traffic pattern and constant radius semi-circles plus straight
line segments for the holding pattern. Figure 2-25 shows a sample autoland trajectory in the
lateral plane with the path to the initial point consisting of several, straight line segments.
Even though both the path to the initial point and the traffic pattern can be represented by
a set of straight line segments, the fact that the traffic pattern trajectory is updated on a
frequent basis, while the path to the initial point remains constant once generated, requires
that the two trajectories are communicated to the FMS differently.

71
 !f
F_F

Expected touchdown point, TD tb

Change active !b
leg here

Initial point, IP

P4 !d F_B
td

!f

Path to the initial point

P3
Holding pattern
Traffic pattern

P2

P1

Figure 2-25: Specification of the autoland trajectory as set of straight line segments and
hold loop

The path planner, that generates the trajectory to the initial point, constructs the path
as a set of waypoints as shown in Figure 2-25 by waypoint P1-P4, with P1 being the position
of the aircraft at the point of autoland initiation and P4 being the initial point of the traffic
patten trajectory. Hence, it is convenient to specify the trajectory to the FMS as a set of
waypoints. For this purpose, so-called track-to-a-fix (TF) legs are used. TF legs require the
specification of two waypoints to fully specify the leg [5]. A complete trajectory is synthesized
by joining as many TF legs as needed. As long as the trajectories built from TF legs are
gap-free, meaning that two subsequent legs have a waypoint in common, the FMS follows
the trajectory without further input required by autoland.

72
 By definition, the FMS will perform a ”fly-by” waypoint, i.e. cutting the corner, if the
course change of the trajectory from one leg to the next is less than 135◦ , and performs a
”fly-over” maneuver, if the course change is greater than 135◦ . For exact calculation of the
length of the trajectory to the initial point, this would need to be taken into account, but
it is assumed that course changes are small and few, so that their effect on the length of
trajectory can be neglected.
For the traffic pattern, TF legs are not practical to use, since updates to the current flight
plan cause a very abrupt reaction, as the guidance function is reset during the re-computation
and capture of the new path. For frequent trajectory updating while in the traffic pattern,
fix-to-a-manual-termination (FM) legs are more suitable. These legs start at a given fix and
are defined by a course that is followed until the operator manually terminates the current
leg and passes a new one [5]. The manual input by the operator is replaced by automatic
action of the autoland software. The fixes are the Initial Point for the downwind leg, F B
for the base leg and F F for the final leg. In order to ensure smooth transition from the
current, active leg to the next leg, autoland needs to ensure, that it passes the new leg at
a distance from the next FM leg fix that is sufficient for capturing the new course without
overshooting the following leg. For 90◦ turns, as is the case for the baseline traffic pattern,
the next leg needs to be injected into the FMS guidance function one turn radius before
reaching the next waypoint.
Even though not strictly enforced for FM legs, the convention that the ”fly-by” waypoint
maneuver is only performed for course changes smaller than 135◦ is adopted here to limit
the amount of corner cutting. For a course change of exactly 135◦ , the next leg would have
to be injected 2.4 times the turn radius before the next waypoint and rapidly increases with
increasing course change from this point. The lead distance to the next waypoint, td , required
for a smooth intercept of the next leg, as a function of change in course, ∆χ, from one leg
to the next, is given by the following function:
�
1 + cos γ
td = r (2.28)
1 − cos γ

73
where γ = 180◦ − ∆χ.
Finally, the holding pattern consisting of two half-circles and variable length straight line
segments, is commanded to the FMS using the hold-to-a-manual termination (HM) leg. The
HM leg requires specification of the direction of the turn, length of the straight line segment,
the location of a hold fix, which is set to the initial point, and specification of the course of
the outbound leg, which is equal to the course of the final leg, χf [5]. When following the
HM leg, the FMS will keep the aircraft in the holding pattern until autoland passes a new
leg. A smooth transition from the holding pattern to the downwind leg is ensured, as long as
autoland does not inject the downwind leg into the FMS before the aircraft has turned onto
the inbound leg of the holding pattern, which is aligned with the course of the downwind
leg.
In addition to the path to the initial point, the holding pattern and traffic pattern, which
are part of the nominal autoland trajectory, autoland includes the ”Fail” trajectory that
consists of a straight line trajectory to an area that is marked as unpopulated on the aero-
nautical charts. For this fail trajectory, autoland uses the Direct-to-a-fix (DF) leg that only
requires specification of the destination waypoint [5]. The FMS then computes the shortest
path to this fix, the ”great circle distance”, and guides the aircraft along this path.

2.2.2 Use of Local and Global Coordinate Systems

In order to simplify the calculations involved in trajectory generation and updating, autoland
uses a local, cartesian North-East coordinate system with the origin located at the runway
threshold, and the runway heading aligned with the positive y-axis, the North axis. The left
graphic in Figure 2-26 illustrates this coordinate system for a left traffic pattern. Specification
of the direction of the traffic pattern plus knowledge of the length of the baseline traffic
pattern is then sufficient to specify the location of the downwind leg fix, base leg fix and
final leg fix as well as the course of each leg within the local coordinate frame. Using this
baseline traffic pattern, the trajectory updating functions can easily derive the location of

74
the new leg fixes and new leg course within the local coordinate frame. Conversion to the
global coordinate frame is achieved by first rotating the traffic pattern trajectory by an
amount equal to the runway heading, and second, linearly translating each leg fix by the
latitude and longitude of the touchdown zone. The location of each leg fix in the North-East
coordinate frame after rotation is calculated using the common rotation formula adjusted for
the difference in convention of positive and negative rotations. Here, a clock-wise rotation
is considered positive.
    
�
Pnorth cos(χr ) − sin(χr ) Pnorth
 =  
�
Peast sin(χr ) cos(χr ) Peast

The right graphic in Figure 2-26 shows the traffic pattern after the rotation by runway
heading, χr . The angle, χr , is negative in this picture and hence produces a counter-clockwise
rotation. By definition, the autoland trajectory generation and updating functions use angles
within range ] − π, π], which requires that the runway heading, χr , returned from the FMS
database is adjusted to this range before performing any calculations with it. χr is expected
to be expressed in true north. Should χr be stored in the FMS with respect to magnetic
north, autoland expects that the correction for magnetic variation is performed internally to
the FMS and the χr returned is expressed in true north.
The latitude and longitude of each leg fix is calculated from the latitude and longitude
of the runway threshold, µ0 and ι0 , and the displacement of the leg fixes, IP, F B and F F
from the origin expressed in latitude and longitude coordinates, ∆µ and ∆ι. A variety of
formulas with various degrees of accuracy exist to perform this calculation. The simplest
assumes the earth is a perfect sphere with radius R = 6, 378, 137m. From the definition of
latitude and longitude on the sphere, ∆µ and ∆ι can be written in terms of ∆N and ∆E as
follows:
∆µ = arctan( R1 )∆N
1
∆ι = arctan( R cos(µ) )∆E

75
 North

North

"r

East
Touchdown
zone threshold
IP East
Rotate by #r (µ0,!0)

Touchdown
"f
zone threshold
(µ0,!0)
F_F’
IP’

F_F
F_B

F_B’

Figure 2-26: Local North-East coordinate system before and after rotation by amount equal
to runway heading

where µ specifies the current latitude.

At 45◦ latitude, the error in neglecting the flattening of the earth towards the poles, re-
sults in an error in estimating ∆µ and ∆ι of less than 0.3%, which would cause an error of
less than 15 ft when specifying the coordinates of the fixes for a 5,000 ft long traffic pattern
leg. This is deemed acceptable. However, the error can easily be removed by using a more
accurate model of the earth.
Using this information, the location of the leg fixes in global latitude and longitude coor-
dinates becomes:
µ = µ0 + ∆µ
ι= ι0 + ∆ι

The resulting leg coordinates can now be used for specification of the leg fixes to the FMS.
The course of each leg is deduced easiest while still in the North-East coordinate frame and
is calculated in a straightforward manner by adding the runway heading to the course of
each leg referenced to the local North-East coordinate system. The FMS expects the course
to be specified with respect to true north within range [0, 2π[, which requires adjustment
from the ] − π, π] range used for the autoland trajectory generation and updating functions.

76
2.3 User Operations

2.3.1 Autoland Initiation

As was shown in Figure 2-1, autoland is envisioned to be initiated by pressing the corre-
sponding button. This task may be performed by a passenger or the partially incapacitated
pilot. For a fully incapacitated pilot, the system would have to automatically recognize that
the pilot is in distress. This automatic engagement could be achieved using a hypoxia sensor
that requests the pilot to acknowledge he/she is still alert if the sensor detects blood oxygen
levels below the set threshold. Autoland would be engaged automatically, if no response is
received within a given time limit. For a first version of the autoland system, this design
proposes to use button press initiation.

2.3.2 Increase Throttle Command

After autoland has detected the engine status, there are two cases, in which autoland would
request the operator to increase the throttle. The command would be given audibly and
visually. In the first case, autoland has detected an engine out scenario, and was unable to
find a runway within range. The aim in this case is to avoid an unnecessary off-field landing
in a situation where the throttle was set below the engine out threshold inadvertently or due
to a planned, steep descent. Second, autoland will request a throttle increase if the engine
power is below the threshold to fly straight and level but above the threshold to land with
the power-off method. In this case, the aircraft is stuck in a situation where it cannot hold
altitude and also cannot land using the autoland final approach scheme. In this situation,
autoland holds the wings level and aims the aircraft at the closest, unpopulated area within
range, while it is waiting for the operator to increase the throttle setting. Figure 2-27 shows
a conceptual view of this command.

77
 Increase throttle
to medium setting

Figure 2-27: Conceptual visual cue for operator to increase throttle

2.3.3 Status Display for Selected Landing Site and Planned Tra-
jectory

After a landing site with corresponding feasible trajectory has been determined, autoland
illustrates the planned trajectory and target landing site on the moving map display as shown
conceptually in Figure 2-28. The aircraft’s movement along the trajectory will be illustrated

Bedford, MA airport within range.

Approaching Runway 11.

Figure 2-28: Conceptional display of target destination and planned trajectory on moving
map

on the display in order to keep the operator informed over the progress.

78
2.3.4 Prepare to Land Tutorial

While the aircraft is following the trajectory to the initial point, autoland will provide the
operator with information on what to expect from the autoland process and what tasks
they will need to perform. The tasks are communicated via the screen with graphics of the
appropriate control input to provide. The first task will be to reduce the throttle to zero
upon arrival at the initial point, if it is not already zero. Only if this task applies, will it be
communicated to the operator. Upon touchdown, the pilot or passenger will be requested to
apply brake pressure in order to assist with the deceleration of the aircraft and keeping the
aircraft aligned with the runway centerline. The location and operation of the brakes will
be illustrated to the operator prior to starting the final approach.

2.3.5 Request to Reduce Throttle at Initial Point

As soon as the aircraft has arrived at the initial point, autoland requests the operator to
reduce the throttle to zero, if it is not zero already. The command is given via audible and
visual cues. A conceptual sketch of the visual cue is shown in Figure 2-29.

Retard throttle

Figure 2-29: Conceptual visual cue for operator to reduce throttle to zero

79
2.3.6 Request to Apply Brakes

Once the aircraft has touched down, autoland requests the operator to apply brake pressure
according to the tutorial shown while flying to the initial point. The command is given via
audible and via visual cues.

80
Chapter 3

System Architecture

The system architecture was derived by decomposing the functionality required to complete
a full autoland mission into individual modules, where each module is responsible for one
specific task only. A full autoland mission consists of selection of a suitable landing site,
trajectory planning to the initial point, tracking of the trajectory to the initial point, gener-
ating, tracking and updating the holding pattern and generating, tracking and updating the
traffic pattern.
Following the decomposition of these functional blocks into a set of tasks, 21 modules
were identified that are each responsible for one specific task. These modules are grouped
into seven components, which are arranged into three layers.
Modules within the same component perform either similar tasks or are each contributing
a part to perform the overall component functionality. Layers represent the level of the
autoland task, with planning tasks being at the top of the hierarchy and trajectory tracking
tasks at the bottom.
The execution of functions and individual modules is controlled by two state machines
running in parallel. Based on the functionality implemented by the components and mod-
ules, and the information required by the state machines, external and internal interfaces
are derived.

81
3.1 Static Decomposition

3.1.1 Autoland Layers

Autoland uses three layers Destination Planning, Trajectory Planning and Trajectory Trans-
lation as shown in Figure 3-2. The Destination Planning layer is responsible for selecting the

Destination Planning

Trajectory Planning

Trajectory Translation

FMS/Autopilot

Figure 3-1: Autoland layers

most suitable landing site given the engine status and achievable range. For this purpose,
the Destination Planning layer interfaces to the FMS databases to query runway information
such as location, length and heading, emergency landing site information and terrain data.
The core of the autoland functionality is provided by a Trajectory Planning layer that
determines the individual legs of the autoland trajectory and updates the trajectory using
the energy management strategies presented in chapter 2.1.9.
The Trajectory Translation layer implements the communication of the trajectory to the
FMS in such a way that the existing trajectory tracking capability of the FMS can be used
without modification. This includes determining the correct timing for passing a new tra-
jectory or a trajectory update to the FMS and translation of the trajectory into a format
that is supported by the FMS.

82
 In addition to that, a user interface layer would be required for the final implementation.
The user interface layer will not be treated in further detail in this thesis, but is an integral
part of the final design, which will require a more thorough treatment from a human factors
perspective than the brief overview presented in chapter 2.3.

3.1.2 Autoland Components

The three layers are further decomposed into components as shown in Figure 3-2.

Destination Planning

Engine Status Detector Landing Site Selector

Trajectory Planning

Trajectory Generator/Updator

Performance Trajectory Translation Energy

Estimator Manager
Auto-
pilot
FMS Interface inter-
face

FMS/Autopilot

Figure 3-2: Autoland system architecture

Components of Layer Destination Planning

The components of layer Destination Planning are Engine Status Detector and Landing Site
Selector. The Engine Status Detector determines whether the engine is operational, has
failed or whether the throttle setting is at an intermediate setting. This information is
used by the Landing Site Selector to determine which search criteria to apply and how to
determine the available range. In case the engine is operational, the range is limited by the

83
available fuel. For case engine out and engine status indeterminate, the range is limited by
the amount of ground range covered over altitude lost.

Components of Layer Trajectory Planning

Layer Trajectory Planning is implemented by components Trajectory Generator, Perfor-

mance Estimator, Energy Manager and Trajectory Updator. The Trajectory Generator is
responsible for creating a new trajectory, either a trajectory to the initial point, the initial
holding pattern or an initial traffic pattern. As the aircraft follows this initial trajectory, the
Performance Estimator monitors the aircraft’s glide ratio and calculates predictions of the
aircraft’s glide performance on the subsequent trajectory segments. The Energy Manager
maintains information on the aircraft’s target energy for the current location and passes the
deviation off the energy target to the Trajectory Updator, if tolerance limits are exceeded.
The Trajectory Updator then shortens or elongates the initial trajectory in order to remove
the energy error.

Components of Layer Trajectory Translation

Trajectory translation is achieved by components FMS Interface and Autopilot Interface.

The FMS Interface is responsible for translating the trajectory returned by the trajectory
planning layer, which is defined in a local, cartesian coordinate system into a global, latitude
and longitude system that can be interpreted by the FMS. Furthermore, the FMS Interface
ensures correct timing of communicating the trajectory updates to the FMS, so that the
aircraft performs smooth transitions from one leg of the trajectory to the next. The Autopilot
Interface sets the autopilot mode depending on the autoland phase. For example, autoland
uses altitude hold while tracking to the initial point (if the engine is operational) and airspeed
hold along the traffic pattern.

84
3.1.3 Autoland Modules

Finally, the components are divided into several modules in order to implement the compo-
nent functionality as shown in Figure 3-3.

Destination Planning

Touchdown Energy
Engine Status Detector Range Calculator Initial Point Calculator
Calculator

Trajectory Planning

Holding Traffic Feasibility

Energy Path Planner Pattern Traffic Pattern
Pattern Monitor
Calculator to Initial Point Generator/ Trajectory
Trajectory
Updator Updator
Generator
Excess Energy
Calculator Landing Site
L/D_0 Trajectory Translation
Prioritizer
Estimator
Energy Curve
Trajectory Manager
to Initial Holding Traffic Autopilot
Flare
L/D_Expected Pattern Pattern Mode
Point Controller Traffic Pattern
Follower Follower Select
Generator Follower Energy Error
Detector

FMS/Autopilot

Figure 3-3: Autoland system architecture

Modules of Component Landing Site Selector

The modules of component Landing Site Selector are the Range Calculator, Touchdown
Energy Calculator, Initial Point Calculator and Landing Site Prioritizer. The Range Calcu-
lator first determines how far the aircraft can go given engine status and the current altitude,
i.e. energy. The Landing Site Prioritizer then searches for landing sites within this range.
For the solutions returned, component Landing Site Selector checks whether the candidate
landing sites are reachable given the altitude of the touchdown zone, using the Touchdown
Energy Calculator. Upon passing this test, component Landing Site Selector calculates the
location of the initial point using the Initial Point Calculator. Subsequently, the Landing
Site Selector attempts to find an obstacle-free trajectory to the initial point that meets the

85
range constraints, using the Path Planner to the Initial Point of component Trajectory Gen-
erator/Updator. If such a trajectory could be found, the Landing Site Selector selects the
landing site and passes the runway heading, location of the runway threshold, the location
of the initial point and traffic pattern direction to the FMS Interface component of layer
Trajectory Translation.

Modules of Component Performance Estimator

The Performance Estimator consists of modules Energy Calculator, L/D 0 Estimator and
L/D Expected Generator, and runs whenever the aircraft cannot hold altitude, i.e. in case
engine out or case engine status ”Indeterminate”. The Energy Calculator computes the cur-
rent energy from altitude and airspeed and the change in energy between sampling intervals.
This sampled change in energy is used by the L/D 0 Estimator to calculate an instantaneous,
and subsequently, averaged L/D0 under consideration of the current wind conditions and
bank angle. The same wind information along with the L/D0 is used by the L/D Expected
Generator to calculate the expected glide ratio, L/Dexp , on the following trajectory segments.

Modules of component Energy Manager

The energy management function is implemented by modules Feasibility Monitor, Excess

Energy Calculator, Energy Curve Manager and Traffic Pattern Energy Error Detector. The
Feasibility Monitor runs while the aircraft is tracking the trajectory to the initial point and
monitors whether the planned destination is still reachable given the current estimation of
L/D0 . If the maximum achievable range becomes less than the remaining distance to go,
the destination is deemed infeasible.
Once the aircraft enters the holding pattern, the Excess Energy Calculator computes
the amount of excess energy the aircraft has over the desired energy to start the traffic
pattern. From knowledge of the current position in the holding pattern provided by the
FMS Interface component if layer Trajectory Translation, the Excess Energy Calculator

86
deduces the amount of excess energy the aircraft will have when next passing the initial
point, which will be used to adjust the length of subsequent loops using the functions of the
Trajectory Generator/Updator component.
Upon generation of the traffic pattern trajectory, the Energy Curve Manager constructs
and maintains the energy vs. range-to-go until touchdown curve from knowledge of energy
dissipation with ground distance travelled, L/Dexp , and the length of the traffic pattern
segments, Li .
The Traffic Pattern Energy Error Detector then uses the mathematical description of this
curve and its tolerance bounds plus knowledge of distance to go until the start of the next
traffic pattern segment, in order to determine whether an energy error exists. If the current
aircraft energy exceeds the tolerance bounds, the Energy Error Detector passes the energy
error on to the Trajectory Generator/Updator component for trajectory adjustment.

Modules of Component Trajectory Generator/Updator

The Trajectory Generator/Updator functionality is implemented by modules Path Planner

to Initial Point, Holding Pattern Trajectory Generator/Updator, Traffic Pattern Trajectory
Generator and Traffic Pattern Trajectory Updator.
The Path Planner to the Initial Point is active during the landing site selection process
in order to determine an obstacle-free path to the initial point at the candidate landing site.
The trajectory is planned in such a way that the aircraft is aligned with the course of the
downwind leg by the time the aircraft reaches the initial point. This allows smooth transition
to either the holding pattern or traffic pattern depending on the available aircraft energy by
the time the aircraft reaches the initial point.
Upon arrival at the initial point, the Holding Pattern Trajectory Generator/Updator cre-
ates a holding pattern with the number and length of hold loops that matches the energy
to dissipate. The holding pattern update process operates in exactly the same way as the
holding pattern generation process, since the Excess Energy Calculator always returns the
expected excess energy on passing the initial point the next time, and not the total excess

87
energy. This was done so that only subsequent hold loops are updated, but not the current
hold loop. This is to conform to the way the FMS accepts HM legs to track a holding pat-
tern. This choice clearly allows non-zero energy error to exist upon exit the holding pattern.
The Traffic Pattern Trajectory Generator takes the current aircraft energy upon exit of
the holding pattern, E/W , the expected glide ratio L/Dexp , on each traffic pattern leg and
the length of the baseline traffic pattern trajectory, and adjusts the length of the baseline
traffic pattern trajectory to match the energy to dissipate until touchdown using the tra-
jectory updating strategy on the downwind leg as presented in chapter 2.1.9. This traffic
pattern is specified in the local, North-East coordinate frame and translated to global lati-
tude and longitude coordinates by the FMS Interface component, for tracking by the FMS.
While the aircraft is tracking the trajectory, the Traffic Pattern Trajectory Updator re-
moves energy errors by using the updating strategy that corresponds to the current leg. The
Trajectory Updator receives information about the current leg from the the Traffic Pattern
Follower and switches updating strategies accordingly. No updating is performed during the
turns from downwind leg to base leg and from base leg to final leg and on the final leg itself.

Modules of Component FMS Interface

The FMS Interface uses modules Trajectory to Initial Point Follower, Holding Pattern Fol-
lower and Traffic Pattern Follower in order to pass on the trajectory generated by compo-
nents Path Planner to the Initial Point, Holding Pattern Trajectory Generator/Updator and
Traffic Pattern Trajectory Generator/Updator to the FMS during the corresponding phases
of the autoland trajectory.
The Trajectory to the Initial Point Follower translates the trajectory returned by the Path
Planner to the Initial Point in local North-East coordinates to global coordinates and passes
this trajectory to the FMS as a set of TF legs.
While in the holding pattern, the Holding Pattern Follower passes the holding pattern
to the FMS as HM legs by specifying the holding pattern fix, the course of the outbound
leg, the turn direction and the length of the straight line segments. The Holding Pattern

88
Follower is also responsible for signaling the FMS to exit the hold at the correct point in
time, so that a smooth transition to the downwind leg is performed.
While in the traffic pattern, the Traffic Pattern Follower passes the traffic pattern trajec-
tory to the FMS as a set of FM legs. The Traffic Pattern Follower passes the legs one at a
time and at a distance from the intersection of two legs that allows a smooth transition from
one leg to the next. Furthermore, the Traffic Pattern Follower inhibits trajectory updates
passed by the Traffic Pattern Updator, while the FMS is capturing a previously updated leg.

Modules of Component Autoilot Interface

The Autopilot Interface consists of modules Autopilot Mode Select and Flare Controller.
Autopilot Mode Select runs continuously and sets the autopilot modes airspeed hold, altitude
hold, and ”Straight and Level” as determined by the two state machines discussed in chapter
3.2. The Flare Controller executes during the autoland flare phase and determines the target
vertical speed based on the aircraft’s altitude above the touchdown zone elevation.

3.2 Autoland Modes and Phases

Autoland uses two state machines to control which of the autoland components and modules
are executing under which circumstances and during which part of the autoland process.
One of the state machines controls the autoland mode, which is directly related to the engine
status. The other state machine controls the autoland phases which is related to the progress
of the aircraft along the autoland trajectory.

3.2.1 Modes

Depending on the engine state, autoland selects either autopilot mode ”Altitude Hold” or
”Airspeed Hold” and turns the performance estimation functionality on or off. Figure 3-4

89
illustrates the autoland modes and the mode transitions.
Upon initiation, autoland sets autopilot mode ”Straight and Level”, in order to abort

Init
-engage autopilot function
"Straight and Level"
Engine status =
"Operational"

Engine status = Engine status =

"Out" "Indeterminate" Engine Operational
entry action:
-engage autopilot mode "Altitude Hold"
during action:
Engine status = none
"Out" exit action:
-disengage "Altitude Hold"
Engine Out
entry action: Engine status =
-start Performance Estimator "Operational Engine status =
-engage autopilot mode "Airspeed Hold" with "Indeterminate"
best glide speed
during action: Engine status =
none "Operational"
exit action: Engine Indeterminate
-stop Performance Estimator entry action:
-disengage "Airspeed Hold" -start Performance Estimator
during action:
none
Engine status = exit action:
"Indeterminate" -stop Performance Estimator

Engine status = "Out" touchdown h<=h_flare

detected

h <= h_flare

Figure 3-4: Autoland modes as determined by engine status

any maneuvering that might have been executed prior to autoland engagement.
If the Engine Status Detector has determined that the engine has failed, autoland enters
mode ”Engine Out” and sets the autopilot mode to ”Airspeed Hold” in order to hold best
glide speed, while the aircraft is following the autoland trajectory at power off. Furthermore,
in mode ”Engine Out” autoland runs the Performance Estimator to determine the glide
capability of the aircraft, which is used determine feasibility of the planned trajectory to
the initial point and to correct energy errors by adjusting the length of the trajectory in the
hold and the traffic pattern.

90
 If the engine power is sufficient for straight and level flight, autoland enters mode ”Engine
Operational” and requests the autopilot to hold altitude in order to maintain altitude up
until reaching the initial point. The Performance Estimator cannot collect information on
the glide performance of the aircraft, while the aircraft maintains altitude and is hence not
active in this mode.
If the engine state is ”Indeterminate”, autoland does not modify the autopilot mode from
the already set ”Straight and Level” mode, but runs the Performance Estimator to obtain an
approximation to the current rate of altitude lost per unit of ground distance travelled, which
in turn is used to determine the aircraft’s available range. The state machine provides for
return from mode engine ”Indeterminate”, upon change of the engine status. This transition
would occur, once the operator followed the command to increase throttle, as given by default
if the engine status is classified as ”Indeterminate”.

3.2.2 Phases

During a nominal automatic landing, autoland traverses phases ”Destination Selection”,

”Approach to Initial Point”, ”Hold at Initial Point”, ”Follow Traffic Pattern”, ”Flare” and
”Rollout” in this order. The corresponding state machine is shown in Figure 3-5. This state
machine runs in parallel with the mode control state machine of the previous chapter.
It is possible to skip phase ”Hold at Initial Point”, if upon arrival at the initial point,
the energy is smaller than the maximum energy allowed to start the traffic pattern and the
power is already off. Besides the nominal path, autoland may go into state ”Fail” in the
unlikely event that no feasible landing site could be found, neither a runway nor a suitable
emergency landing site nor an unpopulated area. This could occur if the aircraft experienced
an engine failure at low altitude above a large metropolitan area. Autoland provides for the
possibility to leave the ”Fail” state and re-attempt a search for a landing site in the event
that the engine status has changed to ”Operational”.
Furthermore, engine state ”Indeterminate”, which cannot be handled by the nominal

91
autoland approach procedure is captured in phase ”Direct-to unpopulated area” if a reachable
unpopulated area has been found. In engine status ”Indeterminate”, autoland does not
attempt to find a runway but rather directs the aircraft straight to an area where it is
expected to encounter the least man-made obstacles and where it is expected to cause the
least damage.

92
 Destination Selection
no feasible landing site found engine status
entry action: (runway or off-field)
-get engine status != "Operational"
during action:
-select feasible landing site Fail
exit action: engine status changed
AND
-display selected landing
engine status = "Operational"
site and planned trajectory
on moving map
engine status changed

trajectory engine status

= "Indeterminate" Direct-to unpopulated area
infeasible
AND entry action:
landing site found landing site found -program FMS with direct-to
landing site type != "runway"
AND waypoint leg (DF leg)
engine status changed during action:
AND Approach to Initial Point nont
engine status = "Operational" entry action: exit action:
-program trajectory to Initial Point into FMS (TF none
legs)
-program hold at Initial Point into FMS (zero-
length straight line segment, HM leg)
during action:
-monitor feasibility of planned trajectory
Initial Point reached
-display braking tutorial
AND
exit action: E<= E_max
Initial Point reached -request pilot to cut throttle if engine status = AND
AND "operational" engine status = "Out"
E > E_max
(max energy at which hold is exited)
Follow Traffic Pattern
entry action:
Hold at Initial Point -generate traffic pattern
entry action: -remove holding pattern from flight plan in FMS
none -pass first leg of traffic pattern to FMS (FM leg)
during action: Initial Point reached during action:
-get engine status AND -remove energy errors by updating traffic pattern
-set # of loops = 1 and length of straight line # loops = 0 -pass updated leg to FMS (FM leg)
segments = 0 if engine status = "Operational" exit action:
-calculate # of loops in pattern and length of -none
straight line segments if engine status = "Out"
-update holding pattern in FMS (HM leg) h<=h_flare
exit action:
-none h>h_flare

Flare
entry action:
-set autopilot mode to "Vertical Speed
#loops != 0 Hold"
OR during action:
Initial Point not -provide exponentially decreasing vertical
yet reached
speed target
exit action: touchdown
not yet detected
-disengage "Vertical Speed" mode

touchdown detected

Rollout
entry action:
-engage "Heading Hold" mode with
runway heading as target
during action:
-request pilot to operate brakes aircraft not
exit action: yet stopped
-disengage "Heading Hold" mode

aircraft stopped

Figure 3-5: Autoland phases determining the lateral trajectory followed by the aircraft

93
3.3 Autoland Interfaces

3.3.1 External Interfaces

In order to perform the functions outlined above, autoland relies on the existing FMS to
provide information from the navigation database and to track a specified trajectory. Fur-
thermore, it relies on the autopilot to provide aircraft state information, such as airspeed and
altitude and to provide limited control of the aircraft’s motion in the vertical plane through
autopilot modes ”Altitude Hold” and ”Airspeed Hold”. In addition to that, autoland expects
to be able to receive weather forecasts from satellite radio. Figure 3-6 shows the interfaces
to the FMS and the autopilot by autoland component that uses it. For a detailed view on

Destination Planning

Engine Status Detector Landing Site Selector

Trajectory Planning

Trajectory Generator/
Updator

Performance Trajectory Translation Energy

Estimator Manager
Auto-
FMS Interface pilot
inter-
face

HM areo runway
TF state
leg mode state charts info data-
legs
state FM select base
DF emergency state
%power leg leg airspeed fuel query
landing site
wind cmd range
info METARS

FMS/Autopilot

Figure 3-6: Autoland external interfaces

which module uses an external input or generates an external output refer to chapter 3.3.4.
On the other end, autoland interfaces to the operator using the display capability of the
integrated flight deck and using audible and visual cues as described in chapter 2.3. However,

94
this chapter focuses only on the autoland interface to the autopilot and the FMS. The exact
external interface to the operator will be determined upon completion of the detailed user
interface design, which is left for future work.
The Engine Status Detector requires the use of external input % power to determine the
engine status. The Performance Estimator uses the wind speed and direction, as well as
the aircraft state, state, estimated by the autopilot to determine the current aircraft glide
ratio and the glide ratio to be expected on subsequent trajectory legs. The FMS Interface
infers the aircraft’s current position from the aircraft state and determines the trajectory to
pass to the FMS to track. While flying to the initial point, the FMS Interface specifies the
track as a set of TF legs. In the holding pattern, the FMS Interface uses HM legs and in
the traffic pattern, FM legs. The DF leg is only used for the ”Fail” scenario where autoland
guides the aircraft directly to an unpopulated area. The Autopilot Interface simply switches
between autopilot modes ”Airspeed Hold” and ”Altitude Hold” and passes the target air-
speed if needed. The Energy Manager uses the current position of the aircraft to determine
the aircraft’s location on the energy vs range-to-go curve in order to deduce potential en-
ergy errors. The Landing Site Selector uses a number of inputs from the FMS’ navigation
database, such as runway info, emergency landing site info and image files of aeronautical
charts in order to compare and select the most suitable landing site. It uses aircraft state
information and fuel range to limit the landing site search region. Furthermore, the Land-
ing Site Selector uses measurements of wind speed and direction for the airport vicinity to
determine the runway direction that yields a headwind on approach. This data is available
from METARs (Meteorological Terminal Aviation Routine Weather Report ) received via
satellite radio. Table 3.1 summarizes the autoland external interfaces to the FMS and the
autopilot.

95
Ident SubIdent Direction Frequency Description

%power - In 1 Hz Estimate of engine power output

wind - In 1 Hz Estimate of wind speed and direction from autopilot
loc In 1 Hz Aircraft’s location in global latitude and longitude
coordinates
hdg Aircraft heading with respect to true north
course Direction of the aircraft’s ground track with respect
state to true north
Va Airspeed in kt
vspeed Aircraft’s current sink rate referenced to the inertial
frame
h Estimate of the aircraft altitude above mean sea level
(MSL)
bank Aircraft bank angle
TF legs - Out triggered Trajectory specified as a sequence of ”Track-to-Fix”
legs
DF leg - Out triggered ”Direct-to-Waypoint” leg
HM leg - Out triggered ”Hold with Manual Termination” type holding pat-
tern
FM leg - Out triggered ”Fix to Manual Termination” leg
mode select - Out triggered Integer identifier specifying the autopilot mode
airspeed cmd - Out triggered Commanded target airspeed in knots
name In triggered Name of the emergency landing site
loc Latitude and longitude of emergency landing site
emerg. LS info
hdg Recommended approach direction with respect to
true north
elevation Elevation of the landing site above MSL
aero charts - In triggered Color Image files of aeronautical charts
fuel range - In triggered Range in nm calculated from amount of initial fuel
and fuel flow
name In triggered String identifier containing the name of the airport
and runway
loc In triggered Latitude and longitude of the location of the runway
runway info
threshold
hdg triggered Runway heading with respect to true north
length triggered Length of the runway including displaced threshold
in ft
dispThres triggered Length of runway absorbed by displaced threshold
elevation Elevation of the runway above MSL
METAR - In 0.5-1 hr Aviation weather report including wind speed and
direction for candidate landing sites
db query - Out triggered FMS database query

Table 3.1: Autoland external interfaces

96
3.3.2 Internal Interfaces

The autoland internal interfaces are used to pass data from one autoland component to
another. Figure 3-7 shows the component interfaces that are active while the aircraft is
tracking the traffic pattern.
The Trajectory Generator/Updator publishes the trajectory in structure ”TP” (Traf-
fic Pattern), which consists of an array of structure ”leg” for each leg of the traffic pat-
tern, TP.leg[], an array of glide ratios TP.L/D exp[] for each segment of the traffic pattern,
TP.L/D exp[], an update flag, TP.updateFlag, and single values specifying the distance to
go until the start of the next segment, TP.R, distance to go until the start of the next leg
TP.L and the number of the traffic pattern segment the aircraft is currently on TP.currSeg.
The array TP.leg[] contains the fix of each traffic patter leg, TP.leg[].fix, the course of each
leg, TP.leg[].course, and the length of each leg TP.leg[].length.
The Performance Estimator uses the information on the course of the traffic pattern legs
and the course change from one leg to the next to determine the expected glide ratios on
each segment of the traffic pattern, TP.L/D exp. The expected glide ratio and the length
of each traffic pattern leg is then used by the Energy Manager to construct the energy vs.
range-to-go curve, that is used to determine whether an energy errors exists.
The FMS Interface uses the leg fixes and courses to synthesize FM legs that it passes to
the FMS one at a time. The FMS Interface keeps track of which segment the aircraft is
currently on and publishes this to the Energy Manger in TP.currSeg. As the aircraft follows
the FM legs, the FMS Interface calculates the range to go until the start of the next segment,
TP.R.
The Energy Manger then uses TP.currSeg and TP.R to locate the aircraft’s position on
the energy vs. range-to-go curve. The Energy Manager thus determines whether an energy
error, ∆E/W , exists and passes the error on to the Trajectory Generator/Updator, if energy
error tolerance limits are exceeded.
The Trajectory Generator/Updator then generates a new trajectory using, ∆E/W , TP.currSeg
and TP.L, to generate a new trajectory that removes the energy error. TP.L is required by

97
the trajectory updating functions, since the distance that the leg fix has to be moved to cor-
rect for the same energy error, increases with decreasing distance from the leg fix. After the
Trajectory Generator/Updator has published the updated trajectory, the process repeats.
Adding the data that is passed while tracking the trajectory to the initial point and

E/W

TP.L/D_exp[]

Performance Estimator
(PE)
TP.leg[].course

Trajectory Generator/
Updator
(TGU) TP.L/D_exp[]

TP.leg[].length
E/W

TP.leg[].fix
TP.leg[].course
TP.L TP.currSeg
Energy Manager
(EM)

FMS Interface
(FMI) TP.currSeg

TP.R

dE/W

Figure 3-7: Autoland component internal interfaces active during phase ”Follow Traffic
Pattern”

while in the holding pattern yields the interface diagram in Figure 3-8. The major addition
are the outputs from component Landing Site Selector that passes the location of the runway
threshold, LS.loc, the runway heading, LS.hdg, the pattern direction, LS.dir, the location
of the initial point, LS.IP, and the desired energy at touchdown, LS.E td, that has been
deduced from the touchdown zone elevation. This information is bundled in structure LS.
In order for the Trajectory Generator/Updator to attempt and create a feasible trajectory
to the selected landing site, the Landing Site Selector passes the available range, which the

98
 Engine Status engine Autopilot
Detector (ESD) status Interface (API)

engine status

E/W
L/D_0
Landing Site Selector LS.hdg
range (LSS)

LS.loc
LS.IP
LS.dir
LS.E_td LS.LS_Ok? LS.loc range
LS.dir
LS.hdg Performance Estimator
LS.IP (PE)
Trajectory Generator/ TP.leg[].course
Updator
(TGU) LS.FP_Ok?

HP.length
HP.#loops TP.leg[].length
TP.leg[].fix HP.dir
TP.leg[].course E/W L/D_0
HP.course
TP.updateFlag HP.fix LS.E_td
TI.length
TP.currSeg TP.L/D_exp[]
TI.leg[].fix HP.L/D_exp[]
TP.L TI.currSeg

Energy Manager
FMS Interface HP.currSeg (EM)
(FMI)

TP.currSeg

TI.L, HP.R, TP.R

dE/W

TP.L/D_exp[]
HP.L/D_exp[]
L/D_0

E/W

Figure 3-8: Autoland component internal interfaces

Trajectory Generator/Updator compares against the length of the best feasible trajectory
it could find. If the range is larger than the length of the trajectory, the Trajectory Gen-
erator/Updator signals the Landing Site Selector that the candidate landing site is feasible
through flag LS.LS Ok.
The added trajectory to the initial point and the holding pattern are specified in struc-

99
tures TI.leg[] and HP respectively. The Energy Manager uses knowledge of which leg of the
trajectory to the initial point the aircraft is currently on, TI.currSeg, the distance to go until
the next waypoint of the trajectory, TI.L and the length of the overall trajectory, TI.length,
to monitor whether the trajectory is still feasible. If not, it signals the Landing Site Selector
using flag LS.FP Ok. Zero wind speed is assumed for the trajectory to the initial point.
Therefore, the feasibility monitor and the Trajectory Generator/Updator use output, L/D 0
from the Performance Estimator.
The Holding Pattern, HP, is defined by the location of the initial point, HP.fix, the turn
direction, HP.dir, the course of the outbound leg, HP.course, the length of the straight
line segments, HP.length, and the number of turns to still complete after the current loop,
HP.#loops. The direction of the holding pattern is opposite of the direction of the traffic
pattern, LS.dir, and the course of the outbound leg is equal to the runway heading, LS.hdg.
The length of the straight line segment, HP.L, and the number of loops, HP.#loops, is sup-
plied by the Trajectory Generator/Updator.
The FMS Interface calculates the range to go until the next segment in the holding pat-
tern, i.e. start of the turn, start of the outbound leg, HP.R, which the Energy Manager uses
to estimate the energy error difference between the aircraft energy at the next pass of the
initial point and the desired energy to start the pattern. This energy error is then used by
the Trajectory Generator/Updator to update the number of loops to remain in the holding
pattern and the length of the straight line segments. Table 3.2 summarizes the autoland
internal component interfaces.

100
Ident SubIdent Origin Destination Frequency Description

engine sta- - ESD API, LSS 1 Hz One of the following: ”Out”=0, ”Operational”=1,
tus ”Indeterminate”=2
range - LSS TGU 1 Hz Achievable range from L/D 0 and available energy
including 30% margin
name LSS HMI triggered String identifier containing the name of the selected
landing site for future display on the user interface
LS hdg LSS FMI, PE triggered Runway heading, i.e. approach direction, with re-
spect to true north
loc LSS FMI triggered Latitude and longitude of the location of the runway
threshold or location of the earliest desired touch-
down point for alternate landing site
IP LSS TGU, FMI triggered Location of the initial point in local North-East co-
ordinates
dir LSS FMI triggered Traffic pattern direction, either ”Right = 1” or ”Left
= -1”
E td LSS EM triggered Desired touchdown energy normalized by weight
from touchdown zone elevation
LS OK TGU LSS triggered Flag indicating whether a trajectory to the candidate
landing site could be found within the available range
FP OK EM LSS triggered Flag indicating whether the current flight plan is still
feasible given the performance of the aircraft
E/W - PE LSS, TGU, 1 Hz Aircraft energy normalized by weight
EM
L/D 0 - PE LSS, TGU, 1 Hz No-wind, wings-level glide ratio estimate
EM
leg[].fix TGU FMI triggered Array containing the waypoints that specify the tra-
jectory to the initial point in local North-East coor-
TI
dinates
leg[].length TGU EM triggered Array containing the length of each leg of the trajec-
tory to the initial point
currSeg FMI EM triggered Counter indicating the current leg of the trajectory
to the initial point the aircraft is on (1 = first leg, ...
n = nth leg)
L FMI EM triggered Distance to go until the next leg fix
length TGU FMI triggered Length of the straight line segment of the holding
pattern
nLoops TGU EM triggered Remaining number of loops after completion of the
HP
current loop
fix TGU EM triggered Location of the holding pattern fix in global coordi-
nates
course TGU EM triggered Course of the outbound leg global coordinates
dir TGU EM triggered Holding pattern direction, either ”Right = 1” or
”Left = -1”
currSeg FMI EM triggered Counter indicating the current segment of the hold-
ing pattern the aircraft is on (1=turn to outbound
leg, 2 = outbound leg, 3 = turn to inbound leg, 4 =
inbound leg)
R FMI EM triggered Distance to go until the start of the next segment
L/D exp[] PE TGU, EM triggered Expected glide ratio on the turn segment and the
straight line segments
leg[].fix TGU FMI triggered Array of trajectory legs specified as set of waypoints
in local North-East coordinates
leg[].course TGU FMI triggered course of each leg in local North-East coordinate sys-
tem
TP
leg[].length TGU EM triggered Length of each traffic pattern segment
currSeg FMI TGU, EM triggered Counter indicating the current segment of the traffic
pattern trajectory the aircraft is on (1 = downwind
leg, 2 = turn to base, 3= base leg, 4 = turn to final
intercept leg (if applicable), 5 = final intercept leg
(if applicable), 6 = turn to final leg, 7 = final leg)
L/D exp[] PE TGU, EM triggered Expected glide ratio on each segment of the traffic
pattern
R FMI EM triggered Distance to go until the start of the next segment
L FMI TGU triggered Distance to go until the next leg fix
updateFlag TGU FMI triggered Flag indicating that a trajectory update has occurred

Table 3.2: Autoland internal interfaces

101
 Ident Value Description

L/D POH 10:1 The glide ratio stated in the Pilot Operating Handbook for an idling engine. For the
SR-22, this is 9.6:1. This value will be set to 10:1 to simplify calculations.
L 5,000 ft Length of a single leg of the traffic pattern. Legs are chosen to be of equal length,
so only one constant is needed. Note: The total length of the final leg is 1,000 ft
longer than the downwind leg and the base leg, but these 1,000 ft constitute the
undershoot margin which places the target touchdown point 1,000 ft past the start of
the touchdown zone. The length of the final leg up until the he start of the touchdown
zone is exactly 5,000 ft, the same as the downwind leg and the base leg.
Lm 1,000 ft Distance consumed by the undershoot margin. Determines the position of the target
touchdown point past the runway threshold
nominal bank an- 30◦ Value increased from the baseline 22.5◦ that the current autopilot uses to reduce
gle, φ0 proportion of the in the traffic pattern spent on capturing course changes.
Vglide 90 kt Assumed best glide speed at which the glide ratio of 10:1 can be achieved.
turn radius, r 1,100 ft Turn radius achieved at Vglide and φ0
L tp,0 15,054 ft Length of the baseline traffic pattern
Va,td 75 kt Desired airspeed at touchdown
vv,td 3.3 ft/s Desired sink speed upon touchdown
L rwy 5,000 ft Desired runway length that is deemed sufficient to perform autoland maneuver if pilot
is operating the brakes
% margin, m 30% Margin applied to calculation of the achievable glide range

Table 3.3: Autoland design constants

3.3.3 Design Constants

In addition to the data carried by the internal and external interfaces mentioned before, the
autoland modules use a number of design constants that do not change from one execution of
autoland to the next. The design constants are chosen so as to match the performance of the
Cirrus SR-22. However, they can be adjusted to match another airplane. Since the design
constants here were chosen with the Cirrus SR-22 in mind, care must be taken in adjusting
not just a single design constant, but to scale other affected constants appropriately as well.
An example would be the length of the baseline traffic pattern legs. If a larger aircraft with
a larger turn radius is to be used, the length of the baseline traffic pattern legs would need
to be increased to maintain the same energy error correction capability while in the traffic
pattern. Table 3.3 summarizes the design constants to be used for the Cirrus SR-22.

102
3.3.4 Complete System Diagram

The full system design as described in the previous chapters, is summarized in Figure 3-9.
This graphic shows the individual autoland modules along with their internal and external
interfaces. Inputs that are derived from design constants, i.e. that are the same for each
execution of autoland, are highlighted in magenta. Inputs that are constant after selection
of a landing site are shown in blue. Interface lines that change value over the course of
the autoland execution are drawn in black. Each module is placed in a colored background
to highlight its affiliation with one component. The exception is module ”Engine Status
Detector”, since it is module and component at the same time. Furthermore, the box color
and box frame of each module was chosen to highlight their point of execution within the
autoland process.

103
 Autoland initiated V_glide Fuel range
runs continuously
L/D_POH
% power Engine Status Autopilot mode % margin L/D_0 runs during Landing Site Selection
Detector (ESD) select E/W
runs during Approach to Initial Point
engine
status Range Calculator
engine (RC) runs during Hold at Initial Point
status
state.loc rwy info runs during Follow Traffic Pattern
desired runway
length, L_rwy emergency landing
site info range

aero charts
image files Landing site
criteria
METARs prioritizer
level aircraft
V_a,td weight, W
LS.hdg
LS.loc
LS.elevation Touchdown
range
Initial Point Energy Calculator
state.loc Calculator
LS.E_td

LS.dir state.loc
LS.loc
TF legs
LS.IP LS.LS_Ok or DF leg
range L_tp,0
Feasibility Monitor
TI.L
Path planner to Trajectory to Initial
L_tp,0 initial point TI.leg[].fix Point Follower TI.currSeg
TI.leg[].length LS.FP_Ok

LS.hdg LS.hdg

aircraft L/D_0 LS.dir

weight W state.h L_m
Energy Calculator r
state.V LS.dir L
LS.hdg LS.E_td
LS.hdg
state.vspeed E/W
state.V E/W Excess Energy HP.R
Calculator HP.curSeg LS.loc
state.bank state.loc
L/D_0 Estimator
state.hdg state.course
r LS.IP
phi
L/D_0 HP.dE/W
state.wind

HP. HM
L/D_(Expected) Holding Pattern HP.length Holding Pattern
L/D_exp[] leg
nom. bank angle, phi Generator Generator/Updator Follower
E/W HP.#loops

TP.L/D_exp[] L_m
L
r

TP.leg[]. Traffic Pattern state.loc

LS.E_td
course Trajectory
(intial) r Generator
TP.leg[].fix, LS.loc
TP.leg[] TP.leg[].course
LS.hdg
v_v,td TP.leg[].length (intial)
TP.leg[].course (initial)
(updated)

r Energy curve LS.E_td

Traffic Pattern
manager
state.vspeed Follower
Flare Controller TP.leg[].length
(updated)
M_u
TP.L/D_exp[] M_l
TP.R
E/W M
vspeed_cmd

Traffic Pattern
FM leg
Energy Error
TP.currSeg
Landing Site Selector Detector

dE/W
Performance Estimator TP.leg[].fix
TP.leg[].course
Trajectory Generator/Updator TP.leg[] (updated)

Energy Manager r Traffic Pattern TP.leg[]

Trajectory
FMS interface Updator

Autopilot interface
TP.updateFlag
TP.L

Figure 3-9: Full system diagram

104
Chapter 4

Simulation Environment

In order to assess the performance achievable with the proposed design of the automatic
landing system, a prototype of the Trajectory Planning layer and the Trajectory Translation
layer for autoland phase ”Follow Traffic Pattern” was implemented in Matlab. Two separate
aircraft models are used to assess this performance. A simple point mass model augmented
with a trajectory tracking controller is used to evaluate the performance of autoland in terms
of touchdown point dispersion. A high fidelity, 6 degree of freedom, rigid body model of the
Cirrus SR-22, supplied by Avidyne, Inc., is used to evaluate autoland’s performance in terms
of flare performance. Flare performance includes the sink rate on touchdown, attitude on
touchdown and tolerance to GPS/WAAS vertical position error.

4.1 Aircraft Model for Evaluation of Touchdown Point

Dispersion

The simulation model for evaluation of touchdown point dispersion uses a point mass model
to simulate the aircraft, and a trajectory tracking controller to imitate the trajectory tracking
capability of the FMS. This model of the aircraft, autopilot and FMS then runs in parallel

105
to a model of the Performance Estimator, and in parallel to a model of the Energy Manager
and the Trajectory Updator, each in a separate thread. Both the Performance Estimator
and the Energy Manager/Trajectory Updator combination are executed once every second
of simulated flight time of the aircraft. Communication between the aircraft/autopilot/FMS
model and the Energy Manager/Trajectory Updator model is achieved via the simulated
FMS Interface component of the Trajectory Translation layer. At each simulation step, the
FMS Interface component publishes the aircraft’s progress along the traffic pattern, which
is then used by the Energy Manager/Trajectory Updator model to determine whether an
energy error exists and hence, whether a trajectory update is required. If a trajectory
update has been performed, the FMS Interface component passes the new trajectory to the
aircraft/autopilot/FMS model for tracking whenever the aircraft is in a state that permits
trajectory updating, i.e. while on the downwind leg or base leg, and while no previous
updates are being captured.

4.1.1 Point Mass Model

The dynamic model for evaluation of touchdown point dispersion consists of a point-mass,
which is free to move in three dimensions. Lateral motion is determined by the aircraft’s
heading, which is set via yaw rate commands from the trajectory tracking controller. No roll
or pitch orientation is defined and hence, their value is set to zero. Motion in the vertical
plane is determined by the glide performance of the aircraft in still air, L/D0 , and the
prevailing wind. The wind speed is referred to by Vw and the wind direction by χw . The
aircraft state consists of 3D position p referenced to the inertial frame, the linear velocity vb
referenced to the body frame, the aircraft orientation with respect to the inertial frame O,
which is determined by the aircraft heading ψ, and the angular acceleration W determined

106
 ˙ The state variables can be written as:
by the heading rate psi.
 
p
 north 
 
p =  peast 
 
pdown

 
Va cos γ
 
 
vb =  0 
 
Va sin γ

where Va is the aircraft’s airspeed and γ = arctan(L/D0−1 ), the aircraft’s flight path angle
referenced to the inertial frame.  
0
 
 
O= 0 
 
ψ
 
0
 
 
W = 0 
 
ψ̇

The rate of change of each state variable is determined by the following equations:
   
cos ψ − sin ψ 0 V cos χw
   w 
   
ṗ =  sin ψ cos ψ 0  vb +  Vw sin χw 
   
0 0 1 0

where Vw and χw are provided by the wind estimator. Since the aircraft is in a trimmed
flight condition, there are no linear body accelerations:
 
0
 
 
v˙b =  0 
 
0

107
By definition,
Ȯ = W (4.1)

Due to trimmed flight, there are also no angular body accelerations:

 
0
 
 
Ẇ =  0 
 
0

With this choice, the aircraft course with respect to the in the inertial frame can be found
from:
ṗeast
χ = arctan( ) (4.2)
ṗnorth
Using this simple point mass model, the flare maneuver cannot be simulated, since the model
has no knowledge of lift and drag, and hence reduction in sink rate and airspeed during a
pitch up maneuver cannot be imitated. However, it is advantageous to use this model
for evaluation of the touchdown point dispersion, since it greatly simplifies the simulation
and allows evaluating the system performance without error sources added from a complete
aircraft/FMS/autopilot model. In order to demonstrate the location of the touchdown point
with the simple point mass model only, the point mass simulation stops at the flare altitude
(70 ft) and positions the touchdown point at the distance from the start of the flare, required
by the nominal flare maneuver, plus any additional distance covered during the flare due to
non-zero wind. The nominal horizontal distance occupied by the flare maneuver is 1,450 ft.
The duration of the maneuver is approximately 9 s. Hence, a 5 kt headwind, for example,
would reduce the horizontal distance covered during the flare by 76 ft from 1,450 ft to 1374
ft.

108
4.1.2 Trajectory Tracking Controller

A trajectory tracking controller is added on top of the point mass model in order to follow
the prescribed trajectory. This controller monitors the distance to go until the next waypoint
provided by the FMS Interface component and initiates a constant turn rate to intercept
the next leg once the distance to go to the next waypoint is equal to the amount of turn
anticipation required to intercept the next leg. In the presence of crosswind, the trajectory
tracking controller commands turn rate corrections to adjust the aircraft heading so that the
aircraft crabs into the wind and uses the horizontal component between its airspeed vector
and the ground speed vector to counteract the wind. The trajectory tracking controller uses
non-linear feedback on crosstrack error using the arctan function similar to the controller
implemented on Stanford’s Stanley, the Stanford Racing Teams entry in the DARPA Grand
Challenge 2005, to determine a desired turn rate, ψ̇ref [12]. The advantage of this feedback
controller is that it saturates the feedback gain towards +π and −π as the crosstrack error
increases to infinity, while maintaining close to linear feedback for small crosstrack errors.
Naturally, the trajectory tracking control is implemented in a different manner in several
control loops on the actual aircraft, but its behavior is assumed to be imitated well with this
single control law. The crosstrack error, e, is found by taking the shortest distance from the
current position to the desired track and adjusting the sign of the crosstrack error according
to the turn direction required to intercept the track. With this convention, the crosstrack
error is negative when the aircraft is located to the right of the track, whereas the crosstrack
error is positive when the aircraft is to the left of the track. In addition to tracking lines, the
trajectory tracking controller ensures that the turn segments are tracked in inertial space,
i.e. it adjusts the turn rate, if wind attempts to blow the aircraft off of the constant radius
turn. The control law is given by:

k2 · e
ψ̇ref = −k1 · arctan( ) + ψ̇track (4.3)
Va

109
where ψ̇track specifies the rate of change of the course of the trajectory to follow. For a
straight line, this value is zero. For a constant radius turn, as would be the case on the the
turn segments of the traffic pattern, the turn rate becomes:

g · tan φ
ψ̇track = (4.4)
Va

Gains k1 and k2 where chosen as 0.5 and 1 for straight line segments respectively, and 1 and 3
for turn segments. With the relatively large gain k1 , it is possible to reach the maximum limit
on the turn rate of the aircraft. The maximum limit is chosen as the turn rate achievable at
a 45◦ bank angle at the current airspeed. For a nominal airspeed of 90 kt, the turn rate limit
rad
becomes 0.22 s
. The controller-aircraft interface is simulated in such a way that a given
turn rate command is followed exactly and without delay. This simplification results in a
better trajectory tracking performance in simulation, than achievable by the actual aircraft.
Any inaccuracy and delay in the aircraft response, which results in the trajectory flown
being shorter or longer than the planned trajectory, results in the aircraft landing long or
short of the target touchdown point. Energy errors ensuing from deviations off the planned
trajectory before the intercept to the final leg can be corrected for by trajectory updating.
Any deviations off the planned trajectory after the intercept to the final leg, however, will
lead to overshoot or undershoot of the target touchdown point by the amount the planned
trajectory differs from the actual trajectory flown. For the purpose of this evaluation, it is
assumed that the touchdown point dispersion due to tracking error is negligible compared
to the effect of GPS error and L/D estimation error. However, this will need to be verified
with the actual aircaft.

110
4.2 Aircraft Model for Evaluation of Flare Performance

4.2.1 High-fidelity Aircraft Model

For evaluation autoland’s touchdown performance with the proposed Flare Controller, a high
fidelity, 6 degree of freedom, rigid body simulation of the Cirrus SR-22, provided by Avidyne,
Inc., was used. This model experiences the aerodynamic forces of lift and drag, propulsive
forces, and gravitational forces. Its motion is controlled via elevator, aileron and rudder
deflection as well as throttle input. The characteristics of the aircraft are captured via non-
dimensional stability and control coefficients. The numerical values of these coefficients is
property of Avidyne, Inc. and are therefore not provided in this description. The coefficients
are used to determine the three dimensional forces, fB , and torques, τB , acting on the aircraft
by dimensionalizing the stability and control coefficients with the current airspeed, altitude
and geometric properties of the aircraft and summing their contribution to the applicable
forces and moments. A detailed treatment of these steps can be found in many texts on
Flight Control such as Airplane and Flight Dynamics and Automatic Flight Controls by J.
Roskam [17].
The aircraft response to the body forces and torques calculated from the stability and control
coefficients is governed by the following equations of motion written in matrix form:

τB = Iω̇B + ωB × IωB (4.5)

fB /m = v̇B + ωB × vB − RT g (4.6)

Ṙ = R × ωB (4.7)
˙ = RvB
∆ (4.8)

where the subscript B denotes that the vector is expressed in body coordinates. vB represents
the linear velocity of the aircraft and ω the angular velocity of the aircraft. R is a 3x3
rotation matrix that transforms a vector expressed in the body frame to the inertial frame.
R can be calculated from knowledge of the angular position of the aircraft’s body frame with

111
respect to the inertial frame and vice versa. RT is the transpose of R and hence converts
a vector expressed in inertial coordinates to body coordinates. g is a 3D vector specifying
the gravitational acceleration with value [0, 0, 9.81m/s2 ]T . I is the 3x3 inertia matrix. ∆
specifies the aircraft position in the inertial coordinate system.

4.2.2 Flare Controller

The flare maneuver is achieved by means of the flare controller, that wraps around a vertical
speed hold controller. The flare controller commands an exponentially decreasing vertical
speed according to the following equation repeated from chapter 2.1.11.

t
v(t) = (v0 − vtd )e− τ + vtd (4.9)

where v0 represents the vertical speed at the start of the flare, vtd the target touchdown
speed at the expected touchdown point, and τ the time constant at which the vertical
speed command is decreased. The flare controller passes the vertical speed target v(t) to
the vertical speed hold controller. The vertical speed hold controller was designed using
the successive loop closure method. It uses feedback on state variables pitch rate, pitch
angle and vertical speed and deduces an appropriate elevator deflection based on the error
of commanded and actual pitch rate, pitch angle and vertical speed. First, the control loop
is closed around pitch rate, then the pitch angle and finally vertically speed. Including
feedback of the pitch angle in addition to pitch rate and vertical speed increases stability
and allows the use of higher gains to speed up the aircraft response. The pitch rate controller
is a simple proportional controller with feedback gain -2.7. The pitch hold controller uses
proportional-integral feedback with gains 1 and 0.8 respectively. Finally, the vertical speed
hold loop wraps around the pitch hold loop with proportional gain -0.03 and integral gain
-0.015. Performance results of this flare controller are given in chapter 5.2.

112
4.3 Wind Model

The wind is modeled as a constant, lateral wind that does not change speed or direction
with time. However, the lateral wind speed and direction changes with altitude, which
allows modeling of the wind shear present in the atmospheric boundary layer as discussed
in chapter 2.1.10. With this approach, at any given altitude, the aircraft’s motion in the
inertial frame can be modeled as the sum of the translation of the wind frame with respect
to the inertial frame and the translation of the aircraft with respect to the surrounding
air. Vertical winds, as would be experienced in microbursts or similar phenomena are not
considered, since these conditions have a strong, negative effect on the achievable autoland
performance, while occurring very rarely. Hence, designing autoland for tolerance to vertical
winds would unnecessarily lower its performance during nominal wind conditions and is
therefore not included. However, any vertical wind has an effect on the measured L/D0 and
would hence be picked up by the L/D0 Estimator and compensated for by the Trajectory
Updator, up to the maximum possible.
Three different wind profiles are used for simulation purposes. The first profile is an
idealized wind shear, where the wind speed reduces linearly from the constant value Vw,0
at 560 ft above the touchdown zone to zero at 0 ft above the touchdown zone. Figure 4-1
illustrates this wind profile. At all times, the wind direction is 180◦ south, so that a runway
orientation of 0◦ north results in a tailwind on the downwind leg and a headwind on the final
leg. Autoland assumes this wind speed profile for calculation of the expected glide ratio on
each leg of the traffic pattern. Hence, if the wind estimator measures Vw,0 at the start of the
downwind leg, this wind speed remains constant up until 560 ft above the touchdown zone
and then decreases linearly to zero, the aircraft will hit the target touchdown point exactly,
assuming no trajectory tracking errors and L/D0 estimation errors have occurred.
The second and third wind models are taken from wind data recorded at the Dallas-Fort
Worth Airport from January 1, 1998 to January 31, 1999 [24]. Profile 2, shown in Figure
4-2, is representative for a condition where the wind speed does not change significantly
with a decrease in altitude. Hence, if the aircraft is flying into a headwind and the L/D

113
 1600 1600

1400 1400

1200 1200
Altitude above TDZ in ft

1000 1000

800 800

600 600

400 400

200 200

0 0
−6 −4 −2 0 2 4 6 0 50 100 150
Wind speed in knots Wind direction in degrees

Figure 4-1: Wind profile for linear reduction in wind speed in boundary layer

Estimator expects the wind speed to reduce linearly, the aircraft will touch down short of
the target touchdown point as will be shown and quantified in chapter 5. The third and
final wind profile shown in Figure 4-3, represents an example of a low-level wind shear,
where the wind changes direction close to the ground. This phenomenon may be observed
in the morning hours when cold air trapped below a warmer layer of air creates an inversion
weather condition.

4.4 Wind Estimator Model

The wind estimator is modeled as a perfect estimator that at any point in time knows the
exact wind speed and direction. Any wind estimation errors present in the actual wind
estimator of the Cirrus SR-22 autopilot will affect the accuracy of the calculation of L/Dexp
on the subsequent traffic pattern legs and result in an overshoot or undershoot of the target

114
 1600 1600

1400 1400

1200 1200
Altitude above TDZ in ft

1000 1000

800 800

600 600

400 400

200 200

0 0
−6 −4 −2 0 2 4 6 −150 −100 −50 0 50 100 150
Wind speed in knots Wind direction in degrees

Figure 4-2: Wind profile for near-constant wind speed in boundary layer

touchdown point. The contribution of the wind estimation error to the touchdown point
dispersion was quantified in chapter 2.1.10 and is hence not included in the simulation, since
the results would be redundant.

4.5 Performance Estimator Model

The Performance Estimator determines the current L/D0 and calculates L/Dexp on the
subsequent legs using knowledge of the wind speed and direction and knowledge of the
course of the subsequent legs to be flown. The implementation of this module follows the
conceptual outline given in chapter 2.1.7. The details of the implementation are given in the
appendix, chapters A.2.2 and A.2.3.

115
 1600 1600

1400 1400

1200 1200
Altitude above TDZ in ft

1000 1000

800 800

600 600

400 400

200 200

0 0
−6 −4 −2 0 2 4 6 −150 −100 −50 0 50 100 150
Wind speed in knots Wind direction in degrees

Figure 4-3: Wind profile for inversion of wind direction in boundary layer

4.6 Energy Manager Model

The Energy Manager determines whether an energy error exists as outlined in chapter 2.1.8
and 2.1.9. Details of the implementation are given in the appendix, chapter A.2.9.

4.7 Trajectory Updator Model

The Trajectory Updator adjusts the length of the trajectory given a non-zero energy error and
the location of the aircraft on the trajectory as described in chapter 2.1.9. Implementation
details can be found in the appendix, chapter A.2.11.

116
Chapter 5

Results

In this chapter, the performance of the autoland prototype is evaluated in terms of touch-
down point dispersion, sink rate on touchdown, attitude on touchdown and tolerance to
GPS/WAAS vertical position error.
Touchdown point dispersion is evaluated using the point mass model described in chapter
4.1, under the three wind profiles as described in chapter 4.3. Uncertainty in wind shear in
the boundary layer is expected to be the largest contributor to touchdown point dispersion
in the final implementation on the aircraft. Other contributors are errors in measurement of
L/D0 and measurement of current wind, and errors in tracking the planned trajectory. The
point mass model removes these error sources by providing perfect aircraft state information
to the Performance Estimator (which results in perfect L/D0 and L/Dexp estimation) and
second, by ensuring near perfect tracking of the planned trajectory. Hence, the effect of wind
can be examined in isolation.
Sink rate on touchdown, attitude on touchdown and tolerance to GPS/WAAS vertical
position error, termed the ”touchdown performance parameters”, are evaluated using the 6
degree of freedom, high fidelity aircraft model, as described in chapter 4.2.1, for a range of
L/D0 . Variations in L/D0 were chosen as test cases, since changes in L/D0 have the strongest
effect on the aforementioned criteria, when compared to the GPS/WAAS vertical position

117
error. However, the effect of the GPS/WAAS error on the touchdown performance parame-
ters is small compared to the effect of variations in L/D0 , as long as the GPS/WAAS error
remains within +/- 6.6 ft (2 m). Lateral wind has no effect on the sink rate on touchdown,
attitude on touchdown or tolerance to GPS/WAAS vertical position error and is therefore
not included in the evaluation of the touchdown performance parameters.

5.1 Evaluation of Effect of Wind on Touchdown Point

Dispersion

This chapter evaluates the effect of uncertainty in wind shear in the boundary layer on the
location of the touchdown point. First, the operation of autoland during tracking of the
traffic pattern is illustrated graphically for the linear wind profile shown in Figure 4-1. This
test will be termed Test Case 1. The same procedure is then performed for the near-constant
wind profile as shown in Figure 4-2 and the inversion wind profile as shown in Figure 4-3.
These tests will be referred to as Test Case 2 and Test Case 3 respectively. Autoland’s
performance under Test Case 2 and 3 is evaluated and discussed, however, detailed graphics
as for Test Case 1 are not provided since the information would be redundant.

5.1.1 Test Setup For Evaluation of Effect of Wind on Touchdown

Point Dispersion

Each test case will be run on a left baseline traffic pattern consisting of downwind leg, base
leg and final leg at right angles to one another and leg lengths of 5,000 ft. The waypoint
locations are given in a local North-East coordinate system with the runway threshold located
at the origin of the coordinate system. The trajectory is oriented in such a way that the
runway heading is 0◦ due north. All distances in the following graphs are given in ft. The

118
baseline waypoint locations are:
   
F FD,east 0 −5, 000
 D,north   
   
 FB,north FB,east   −5, 000 −5, 000 
TP.leg[].fix = 

=
 


 FF I,north FF I,east   −5, 000 0 
   
FF,north FFe ast −500 0

The resulting test trajectory is shown in Figure 5-1. The point marked with the red star

2000

Baseline Trajectory

1000

−1000

−2000

−3000

−4000

−5000

−7000 −6000 −5000 −4000 −3000 −2000 −1000 0 1000 2000

Figure 5-1: Left traffic pattern trajectory used to test the performance of the autoland
software prototype

indicates the target touchdown point. The two black dots indicate the end of the overshoot
and undershoot margin respectively. The no-wind, wings-level glide ratio L/D0 is taken to
be 10:1, which results in a glide ratio of 8:1 during a 30◦ banked turn.

119
5.1.2 Illustration of Test Case 1 for Evaluation of Effect of Wind
on Touchdown Point Dispersion

Test Case 1 operates under the linear wind profile shown in Figure 4-1. The wind speed is
5 kt in direction 180◦ south, decreasing linearly to zero from 560 ft. This is the exact wind
profile autoland assumes. Hence, there is no error in estimating the expected glide ratio
for each leg L/Dexp . Hence, assuming no trajectory tracking errors, the aircraft would land
right on the target touchdown point.
The aircraft is started at altitude 1950 ft, approximately 100 ft higher than required to
follow the baseline traffic pattern. This results in a first trajectory update right at the start
of the traffic pattern as shown in Figure 5-2. The Trajectory Updator deduced the amount

2000

1000 Baseline trajectory

Updated trajectory

aircraft start altitude

−1000 100 ft too high

−2000

−3000

−4000

−5000

−7000 −6000 −5000 −4000 −3000 −2000 −1000 0 1000 2000

Figure 5-2: Traffic pattern after update from 100 ft altitude error

to move the base leg fix and final leg intercept fix, d, from the distance to go until the
base leg fix, L, the current angle of the downwind leg to the base leg, αp and the average
glide ratio expected on the remainder of the trajectory, L/Dav . d becomes 500 ft, which
�
results in a shift of the base leg fix by 354 ft south and 354 ft west to FB,north = −5354 and

120
 �
FB,east = −5354. The final leg intercept fix was moved 534 ft south to FF� I,north = −5354
and FF� I,east = 0. The energy vs. range-to-go curve before the update is shown in Figure
5-3. The single square indicates the current energy level of the aircraft at the expected

2000

1950
1800

1900
1600
1850

1400
1800

1200 1750

1000 1700

1650
800

1600
600
1550

400
1500

200
0 5000 10000 15000 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5
4
x 10

Figure 5-3: Energy vs. range-to-go curve be- Figure 5-4: Energy vs. range-to-go curve be-
fore first update fore first update, enlarged

range-to-go until touchdown. The blue curve is the desired energy level, whereas the red
curve and green curve represent the upper and lower energy boundaries respectively. After
the first update, the trajectory is elongated in such a way, that the current aircraft energy at
the new range-to-go until touchdown coincides with the blue curve. This is shown in Figure
5-5. Close inspection shows that the energy level of the aircraft has remained constant at
1950 ft, while the range-to-go has been increased from 14,000 ft to 15,500 ft.
As soon as the aircraft intercepts the base leg, the aircraft’s L/D0 is dropped by 5%.
This may occur, for example, due to unexpected flap extension. The drop in L/D0 requires
a shortening of the trajectory. Since this drop occurred sufficiently early on the base leg,
the full energy error resulting from this drop, can be corrected. The updated trajectory
is shown in Figure 5-7 in red. The blue line on the graph indicates the actual trajectory
flown by the aircraft so far. The corresponding energy vs. range-to-go curves before and
after the update are shown in Figure 5-8 and 5-10. Before the update, the aircraft energy
is 64 ft below the energy level required. Taking into account the distance to go until the
final leg intercept fix, L, the angle of the currently planned base leg to the final intercept

121
 2000

1800 1950

1600 1900

1850
1400

1800
1200

1750
1000

1700
800

1650
600

1600
400

1550
200
0 2000 4000 6000 8000 10000 12000 14000 16000 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5 1.55 1.6
4
x 10

Figure 5-5: Energy vs. range-to-go curve after Figure 5-6: Energy vs. range-to-go curve after
first update first update, enlarged

2000

1000

−1000

−2000

−3000

−4000
Flaps extended

−5000

−7000 −6000 −5000 −4000 −3000 −2000 −1000 0 1000 2000

Figure 5-7: Traffic pattern after second update due to reduction in L/D0

leg, αp and the average glide ratio expected on the remainder of the trajectory, L/Dav , the
Trajectory Updator deduced that the final leg intercept fix should be moved by a total of
d =540 ft, which corresponds to a change in the North and East coordinate of 380 ft to
FF� I,north = −4975 and FF� I,east = −380. Since no previous updates have been performed on
the base leg, dtotal = 0.
Due to the wind behaving exactly as expected and the trajectory tracking controller

122
 1800

1600
1600

1400 1500

1200
1400

1000

1300
800

1200
600

400 1100

200
0 2000 4000 6000 8000 10000 12000 5000 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000

Figure 5-8: Energy vs. range-to-go curve be- Figure 5-9: Energy vs. range-to-go curve be-
fore second update fore second update, enlarged

2000
1450

1800
1400
1600

1350
1400

1200 1300

1000 1250

800
1200

600
1150
400

1100
200
0 5000 10000 15000 5500 6000 6500 7000 7500 8000 8500 9000 9500

Figure 5-10: Energy vs. range-to-go curve af- Figure 5-11: Energy vs. range-to-go curve af-
ter second update ter second update, enlarged

following the prescribed trajectory precisely, the aircraft finally touches down at the target
touchdown point with an error of <10 ft. Figure 5-12 summarizes the initially planned
trajectory, the trajectory after the first update and after the second update, and the actual
trajectory, the aircraft has flown.

123
 2000

Baseline trajectory
After 1st update
1000 After 2nd update
Actual trajectory flown

−1000

−2000

−3000

−4000

−5000

−6000
−6000 −5000 −4000 −3000 −2000 −1000 0 1000

Figure 5-12: Trajectory before and after each update and actual trajectory flown

124
 Test Case 1: Linear
Test Case 2: Test Case 3: Inversion
wind profile
Near-constant wind wind profile
profile
Initial energy error, ∆E/W1 100 ft 135 ft 109 ft
Distance base leg fix was 500 ft 690 ft 544 ft
moved, db
Energy error after change in -49 ft -57 ft -62 ft
L/D0 , ∆E/W2
Distance final leg intercept fix -435 ft -534 ft -564 ft
was moved, df i
Horizontal distance covered 1,450 ft 1,379 ft 1,569 ft
during flare, xh
Overshoot/undershoot of the 7 ft -398 ft +75 ft
target touchdown point

Table 5.1: Summary of test results

5.1.3 Summary of Effect of Wind on Touchdown Point Dispersion

Table 5.1 summarizes the touchdown point dispersion of the autoland prototype during Test
Case 1 to 3. The energy errors, ∆E/W1 and ∆E/W2 incurred at the start of the test run due
to excess altitude, and on the base leg due to the change in glide performance, are included
for reference. Furthermore, the change in location of the leg fixes, db and df i is given to show
the magnitude of the changes in the trajectory for a given energy error. As expected, df i for
updates on the base leg is greater, than db for updates on the downwind leg for the same
energy error. Furthermore, the horizontal distance covered during the flare, xh , is included
to show to effect of wind on the flare distance.
Test Case 1 was discussed in the previous section. In Test Case 2, the aircraft landed 398
ft short of the target touchdown point. This is due to the fact, that autoland assumed that
the wind speed would reduce from 5 kt to 0 kt, as the aircraft descended along the final leg.
Since the wind speed only reduced to 4.5 kt, however, the aircraft still experienced a tailwind
of magnitude 3.9 kt upon touchdown. Refer to Figure 4-2 for verification of wind speed and
direction. Stepping through the calculations outlined in chapter 2.1.10 on determining the
glide ratio estimation error, results in L/Derr = +6.4%. For a 6,000 ft long final leg, this
L/Derr results in an undershoot of the target touchdown point of -394 ft, which is close to

125
the value determined from simulation. The ground distance covered during the flare is less
than the nominal flare distance by 71 ft, since the aircraft still experienced a headwind of
3.9 kt for the duration of the flare.
In Test Case 3, the aircraft landed 75 ft beyond the target touchdown point. Here, the
wind speed initially reduced less than expected over the length of the final leg, but then
abruptly changed direction from a headwind to a tailwind at 200 ft of altitude. This change
in direction suddenly increased the glide ratio of the aircraft and changed the landing that
would have been short, into a landing that was slightly long. The distance covered during
the flare in this case was longer than nominal by 119 ft, since the aircraft experienced a
tailwind of 2 kt of magnitude for the duration of the flare.

5.1.4 Discussion of Effect of Wind on Touchdown Point Dispersion

The autoland prototype met the performance specifications for Test Cases 1 and 2 by keep-
ing the touchdown point dispersion within the unilateral tolerance limit of -400 ft for errors
in predicting boundary layer wind shear in a headwind case. For an approach in a tailwind,
the error margin for touchdown point dispersion would have been +400 ft.
In Test Case 3, autoland assumed an approach into a headwind. The change in wind
direction, however, changed this case into an approach into a tailwind that caused an over-
shoot of 75 ft, which was not accounted for by the unilateral tolerance bound for wind shear.
However, this small overshoot is acceptable, since the length of the runway easily absorbs
this magnitude of error. If Test Case 3 would have been an approach with a tailwind, how-
ever, the change in wind direction would have caused an undershoot, that was not accounted
for. If all other error tolerance bounds for GPS vertical position error, and L/D0 and wind
measurement error, were exhausted at their undershoot limit, the aircraft would touch down
short of the runway threshold, which should be avoided. Hence, the currently unilateral
tolerance bound for error in wind shear prediction should be extended to contain margin in
the opposite direction as well, in order to be able to absorb changes in wind direction.

126
 Since no GPS error and no performance estimation error due to measurement error was
included in the simulation, the additional overshoot and undershoot margin of +/- 600 ft
was not used.
In the actual aircraft/FMS/autopilot system, errors in estimating wind speed and direc-
tion, and L/D0 will naturally occur. This effect is expected to produce a touchdown point
dispersion of less than +/-300 ft. However, this will need to be verified in practice.
Inevitably, there will also be a non-zero GPS error, which causes the aircraft to touch
down sooner or later during the flare maneuver. The overshoot and undershoot caused by
the GPS error is accounted for by bilateral tolerance bounds of +/- 300 ft. The capability
to absorb the GPS error in terms of tolerance on touchdown airspeed is discussed in the
following chapter.

5.2 Evaluation of Flare Performance

In this chapter, the performance of autoland’s flare controller is evaluated with respect to sink
rate on touchdown, attitude on touchdown and tolerance to GPS/WAAS vertical position
error for different L/D0 . In addition to that, the horizontal distance travelled during the
flare maneuver is examined.
In the absence of a vertical wind component, non-zero wind does not change the sink
speed of the aircraft. Hence, the flare controller acts on an aircraft with a specific L/D0 in
the same manner, regardless of whether wind is present or not. Non-zero wind changes the
distance the aircraft travels over ground during the flare maneuver. However, this effect has
already been included in the previous chapter and hence zero wind is used for this test.

5.2.1 Test Setup for Evaluation of Flare Performance

The performance of the flare controller is tested using the high fidelity, 6 degree of freedom,
rigid body simulation using three different L/D0 : 7, 10 and 13 representing Test Case 1, 2

127
and 3 respectively. The three different L/D0 were achieved by adjusting the zero-lift drag
coefficient, CD,0 . An L/D0 of 10 is expected to be the nominal glide ratio, 7 the low end of
the glide ratios to be expected, and 13 the upper end. In each case, the aircraft descends
from an altitude 200 ft in its trim condition at airspeed of 90 kt and zero thrust. When
reaching 70 ft altitude, the flare maneuver is initiated and the flare controller turned on.
The test setup is illustrated in Figure 5-13. For a detailed description of the flare maneuver,

State Machine Vv,cmd Vv,err Vv, h

(Vv,cmd= Vv,0 for h>70 ft + PI-vertical speed !e Avidyne SR-22
Vv,cmd= Vv,flare for h<=70 ft) - hold controller Simulink Model
)
Vv
h

Figure 5-13: Test setup for performance evaluation of flare controller

refer to chapter 2.1.11.

5.2.2 Illustration of Test Case 1 for Evaluation of Flare Perfor-

mance

The following test run shows the simulated aircraft’s behavior during the flare maneuver for
Test Case 1, with an L/D0 of 7, the hardest case for the flare controller to handle. This
is due to the energy being lowest for an aircraft with L/D0 of 7, compared to an L/D0 of
10 or 13, if all else is equal. Yet, the flare controller needs to slow down the sink speed of
the low energy aircraft to the same target sink speed of 3.3 ft/s as for the higher energy
aircraft. This is achieved by trading airspeed for sink speed. However, airspeed can only
be traded up until the airspeed has reduced to the stall speed. Below this airspeed limit,
the aircraft would experience a sudden decrease in lift, resulting in a sudden increase in sink
rate, opposite the desired effect. However, as discussed in chapter 2.1.11, the parameters of
the flare controller were chosen so that sufficient stall margin is available even in the worst
case scenario considered.

128
 Figure 5-14 shows the sink speed command given by the flare controller and the aircraft’s
response. This simulation assumes zero vertical GPS error, but considers the effect of touch-
ing down 6.6 ft (2 m) earlier or later during the summary of the results in the following
chapter. The sink speed command is shown via the solid line, whereas the aircraft response

Vv,0= 20.8 ft/s

Vv,cmd

vertical 4 Vv
speed in
m/s
3

Vv,td= 3.3 ft/s

0
0 2 4 6 8 10 12 14 16

time t in s
Time offset: 0

Figure 5-14: Commanded sink rate and actual aircraft sink rate during flare maneuver

is shown via the dotted line. The flare maneuver is started after 5.5 s, after the aircraft has
sunk to 70 ft altitude. From this point on, the flare controller commands an exponentially
decreasing sink speed from 20.8 ft/s to 3.3 ft/s. The aircraft follows the sink speed command
reasonably well and manages to establish a sink rate of 3.8 ft/s upon touchdown. The swing
in aircraft sink speed at the bottom of the exponential sink speed command is due to the
aircraft entering the ground effect, which is turned on as a step input at 33 ft of altitude.
Figure 5-15 shows the aircraft’s motion in the vertical plane during the flare maneuver. At
around 40 ft, the decrease in sink rate becomes noticeable. The descent into the ground
effect at 33 ft of altitude is again noticeable as a small bump in the vertical profile.
Figure 5-16 shows the horizontal and vertical position, as well as the horizontal and ver-
tical speed of the aircraft during the flare maneuver. The plot of horizontal speed vs. time
shows that the aircraft touched down with an airspeed of 75 kt, which is above the zero flaps
stall speed of 73 kt [8].

129
 200

180

160

140

120
altitude in ft

100

80
ground effect starts (33 ft)
60

40

20

0
0 500 1000 1500 2000 2500
horizontal distance in ft

Figure 5-15: Altitude vs. horizontal distance travelled during flare maneuver

2500 100

2000
x position in ft

90
V in kts

1500 Vtd= 75 kts

1000 80

500
70
0
0 5 10 15 20 0 5 10 15

200 0

−5
150
z position in ft

Vv,td= 3.8 ft/s

Vz in ft/s

−10
100
−15
50
−20

0 −25
0 5 10 15 20 0 5 10 15 20
time in s

Figure 5-16: Change in aircraft position and velocity during flare maneuver

Figure 5-17 shows the evolution of the angle of attack, α, the flight path angle γ, and
the pitch angle θ during the flare maneuver. For an L/D0 of 7 with zero wind, the aircraft
descends at an angle of −8.1◦ . By the time of touchdown, γ has increased to −1.5◦ , which is

130
 0.2

alpha in rad
0.15

0.1
0 2 4 6 8 10 12 14 16
0

gamma in rad
flare maneuver
initiated
−0.1

−0.2
0 2 4 6 8 10 12 14 16
0.2

pitch in rad 0

−0.2
0 2 4 6 8 10 12 14 16
time in s

Figure 5-17: Evolution of angle of attack, flight path angle and pitch during flare maneuver

the flight path angle that determines the overshoot or undershoot of the target touchdown
point, for a non-zero GPS vertical position error. With this flight path angle, the assumed
GPS error of +/- 6.6 ft (2 m), produces an overshoot and undershoot of +/- 300 ft as pre-
viously illustrated in chapter 2.1.10. The angle of attack required to create the lift that
balances the aircraft weight at this flight path angle is 6.8◦ . During the flare maneuver, α
increases to 9◦ , which provides good margin to the stall angle of attack of 14◦ . The aircraft’s
pitch angle is slightly negative at the start of the flare, meaning, the aircraft is in a slight
nose-down attitude. Due to the increase in flight path angle and decrease in airspeed during
the flare maneuver, however, the pitch angle increases to a pitch-up attitude of 7.5◦ . This
attitude is deemed sufficient to avoid nose-wheel barrowing on touchdown.

5.2.3 Summary of Flare Performance Test Results

Table 5.2 summarizes the flare controller’s touchdown performance for L/D0 of 7, 10 and 13
in terms of the touchdown sink rate Vv,td , the touchdown airspeed, Va,td , the pitch attitude,
θtd , the available stall margin, m and the horizontal distance travelled during the flare, xh .

131
 L/D0 = 7 L/D0 = 10 L/D0 = 13
(CD,0 = 0.05) (CD,0 = 0.02) (CD,0 = 0)
Pitch angle at start of flare, θ0 −1.3◦ 1.1◦ 2.4◦
Sink rate at start of flare, Vv,0 20.8 ft/s 15.5 ft/s 11.6 ft/s
Pitch angle at touchdown, θtd 7.5◦ 7.2◦ 6.7◦
Sink rate at touchdown, Vv,td 3.8 ft/s 3.1 ft/s 2.9 ft/s
Airspeed at touchdown, Va,td 75 kt 78 kt 81 kt
stall margin, m 8.25 ft (2.5 m) 13.2 ft (4.0 m) 23 ft (6.9 m)
Horizontal distance covered dur- 1,450 ft 1,508 ft 1,566 ft
ing the flare, xh

Table 5.2: Summary of flare performance test results

The pitch attitude, θ0 , and sink rate, Vv,0 , at the start of the flare are provided for reference.

5.2.4 Discussion of Flare Performance Test Results

The flare controller met the required touchdown performance in each case. As discussed in
chapter 2.1.11, the sink speed on touchdown should remain within 1 ft/s to 6 ft/s to avoid
damage to the aircraft and passengers. The flare controller managed to keep this parameter
within +/- 0.5 ft/s from the target 3.3 ft/s. Each time, the aircraft landed with a pitch up
attitude of > 5◦ , which is deemed sufficient to prevent the nose wheel from touching down
first. The airspeed remained above the stall speed of 73 kt, and provided a stall margin of >
6.6 ft (2 m) in each case to absorb the expected vertical GPS error. For a positive GPS error,
i.e. the aircraft touches down 6.6 ft (2 m) higher in altitude than expected, inspection of
the plot of vertical speed vs. time and flight path angle vs. time from chapter 5.2.2 reveals,
that the touchdown sink speed and flight path angle would only be slightly higher than in
the non-zero GPS error case and is hence well within the the acceptable region.
Due to the differences in sink speed at the start of the flare for each L/D0 , the horizontal
distance travelled during the flare is different in each case. The aircraft with the worst
glide ratio of 7:1 touched down 50 ft earlier than the aircraft with the nominal glide ratio
of 10:1. Conversely, the aircraft with the best glide ratio of 13:1, touched down 58 ft later
than the nominal aircraft. Compared to the touchdown point dispersion from errors in

132
expected glide ratio, L/Dexp , as discussed in the previous chapter, changes in L/D0 add only
little touchdown point dispersion. Hence, the design choice is to absorb this small effect
by assuming a nominal flare length of 1,450 ft, i.e. for the worst assumed glide ratio. Any
deviation off this flare length then results in a small overshoot of the target touchdown point,
which is tolerable.

133
134
Chapter 6

Conclusion and Future Work

6.1 Conclusion

The design for an autoland system for general aviation aircraft presented, aims to address
the emergency scenarios of partial or complete pilot incapacitation and loss of power. The
scope of the proposed concept encompasses automatic landing site selection and guidance
to the selected landing site in addition to the usual scope of an autoland system, which
only covers the final approach to land. The challenge arising from lack of auto-throttle and
lack of radar altimeter, as typically encountered in general aviation aircraft, was alleviated
by performing the final approach at power-off and placing the target touchdown point, so
as to provide overshoot and undershoot margin for expected aircraft performance and GPS
position errors.
The theoretical analysis presented and performance results gathered in simulation show
that an automatic landing of a general aviation sized aircraft is possible on a 5,000 ft runway,
if the error in L/Dexp prediction from measurement errors is kept within +/- 5% and the
error from predicting wind shear is kept within -6.5% for a headwind approach and +6.5%
for a tailwind approach. The vertical GPS error should not exceed +/- 6.6 ft (2 m) to remain
within the allotted overshoot/undershoot margin and the stall margin provided by the flare

135
controller.
Touchdown point dispersion from errors in trajectory tracking performance were deemed
to be negligible. However, any deviation off the planned trajectory, starting from the turn
onto the final leg, will contribute to overshoot or undershoot of the target touchdown point
by the amount the actual trajectory flown differed from the trajectory planned.
Assuming perfect trajectory tracking and an approach into a headwind, the earliest touch-
down would occur at the runway threshold, whereas the latest touchdown would be 1,600 ft
past the runway threshold, if each error source caused the maximum positive and maximum
negative allocated error respectively. On a 5,000 ft runway, this leaves 3,400 ft of rollout
distance, for a range of rollout distance from 1,000 ft - 2,000 ft required by a trained pilot
for the Cirrus SR-22 [8]. Given that the system is intended to be used by an untrained
passenger in case of pilot incapacitation, the additional margin is deemed necessary for a
successful landing by an untrained pilot.

6.2 Future Work

The simulated autoland performance results appear promising, however, the results would
need to be verified during flight testing. Furthermore, performance enhancement of the
current design may be necessary in order to compensate for difficulties with tight trajectory
tracking, or in order to land on runways shorter than 5,000 ft. The performance of the
autoland design presented can be improved in two ways: via operational means by adding
more complex trajectory updating, or via removal of one or more hardware constraints. Both
means are briefly discussed.

136
Performance Enhancement through Addition of More Complex Trajectory Up-
dating

The current trajectory updating methodology in the traffic pattern allows for trajectory up-
dating only on the downwind leg and base leg using two straightforward schemes. During
turns and on the final leg, trajectory updating is inhibited. However, for maximum perfor-
mance, trajectory updating would need to be possible at all times. This could be achieved
by incorporating the capability to change the course of a previously updated traffic pattern
leg, while the aircraft is in the process of intercepting the updated leg. Furthermore, more
complex maneuvers that allow a change in turn radius or turn direction, while executing a
maneuver could be added to control to a desired energy vs. range-to-go state at all times.
The current design locks the decision about the shape of a maneuver at the start of the
maneuver and only allows updates after the completion of the maneuver.

Performance Enhancement from Removal of Hardware Constraints

The current design was developed with the hardware restrictions typically found on general
aviation aircraft in mind. These constraints are: lack of automatic throttle control, lack of
auto-rudder, lack of auto-brakes and lack of radar altimeter. However, if these limitations
were removed, the performance of autoland could be improved. Figure 6-1 repeats Figure 1-1
from chapter 1.4 with the added automatic control lines shown as dashed lines. Also, auto-
flaps are included in the figure. Flaps are not typically driven by a conventional autoland
system. However, addition of this capability would offer automatic glide ratio modulation,
which would be of use to the autoland system proposed.
If automatic throttle control were to be added to the the aircraft, the limitation on the
achievable flight path angles and airspeed would be removed. Hence, any flight path angle
and airspeed within the performance envelope of the aircraft could be flown, meaning the
standard −3◦ flight path angle could be flown and power adjustments could be used to hold
the aircraft on this glide path even under changing wind and aircraft performance.

137
 Operator

autoland
status,
commands
aileron

elevator
Autoland
throttle

mode
rudder
navigation lateral
select aircraft brakes
data trajectory state

flaps
Flight Management System

roll steering commands

Autopilot

aileron throttle brakes

elevator rudder flaps

Aircraft

Figure 6-1: Autoland integration with its environment

Addition of automatic rudder, would provide two benefits. First, it allows for the execu-
tion of a ”de-crab” maneuver during a crosswind landing, which aligns the nose of the aircraft
with the runway centerline just before touchdown. Second, it adds the possibility to control
the steepness of the glide path without adding airspeed, by manipulating the aerodynamic
efficiency of the aircraft. This would be achieved by performing a deliberate slip, in which
the aircraft is banked using the ailerons, and the resulting tendency to turn counteracted by
the application of opposite rudder. The amount of flight path angle control available with
this maneuver depends on the rudder authority available. The rudder needs to counteract
the horizontal component of the lift vector that tends to pull the aircraft into a turn, in
order to maintain the current aircraft heading. As soon as maximum rudder deflection is
reached, further steepening the bank angle, results in the aircraft turning.

138
 Automatic flap deployment could be used to deliberately modulate the aircraft’s glide
ratio. Since extended flaps lower the glide ratio, this control input could be used in a similar
manner as the slip via the automatic rudder. In this respect, automatic flaps are redundant
to the automatic rudder. However, automatic flaps could be used to extend the glide ratio
modulation range of the rudder, whenever rudder authority has been exhausted.
Addition of a radar altimeter would remove the touchdown point dispersion incurred from
the GPS vertical position error, since the measurement error of the radar altimeter is negligi-
ble compared to the GPS position error. This would also allow autoland to control to a softer
touchdown sink rate than 3.3 ft/s, since no airspeed margin is required for a potentially late
touchdown due to GPS error.

139
140
Appendix A

Module Descriptions

A.1 Modules of Layer Destination Planning

A.1.1 Engine Status Detection

Inputs

• % power, P: estimation of engine power derived from manifold pressure, indicated air
speed, outside air temperature, pressure altitude, engine speed, and fuel flow within
the Cirrus SR-22 and made available on the autopilot data bus by the integrated flight
deck

Outputs

• engine status: engine status, 0 = ”Out”, 1 = ”Operational”, 2 = ”Indeterminate”

Trigger

Initiation of autoland

141
Description

The engine status will be classified as either ”Engine Operational”, ”Engine Out” or ”Inde-
terminate”. The engine status is deduced to be ”Engine Operational”, if the engine output
power is equal to or higher than the power required to maintain straight and level flight. This
setting is denoted by pu . For the SR-22, this value is about 20%. The engine is deduced to
be ”Out”, if power output is smaller than the maximum power setting at which a successful
power-on approach can be performed. This value is referred to by pl . For the SR-22, this
is about 5%. In case the power setting is in between pu and pl , the engine status is set to
”Indeterminate”.

The logic for output ”Engine Status” becomes:



 0, if P < pl


Engine Status = 1, if P ≥ pu



 2, if p ≤ P < p
l u

A.1.2 Range calculator

Inputs

• Engine status: engine status output from Engine Status Detector

• Fuel range, Rf : achievable range in nautical miles calculated internally to the SR-22
from amount of initial fuel (set manually) and fuel flow

• E/W: current aircraft energy normalized by weight (input used if engine status =
”Out” or ”Indeterminate”)

• L/D POH: expected glide ratio of the aircraft for an idling engine, design constant

• L/D 0: estimated no-wind, wings-level glide ratio of the aircraft from the Performance
Estimator

142
 • LS.E td, Etd /W : desired energy on touchdown normalized by weight, design constant

• % margin, m: percent margin to be used over maximum achievable range as safety

factor

Outputs

Range, R: achievable range radius

Trigger

called by higher level component, Landing Site Selector

Description

The Range Calculator uses input ”engine status” to decide whether to route input Fuel
range, Rf , or a calculated glide range, Rg to output R. If the engine is operational, R is
assigned Rf . In the remaining cases, R is assigned Rg :

 R , if engine status = ”Out”
f
R=
 R , if engine status =
� ”Out”
g

The glide range, Rg including margin, m can be calculated from the simple formula:

1
R= (E/W − Etd /W ) · L/D (A.1)
1+m

where L/D is either the L/D 0 returned from the Performance Estimator or the average
expected glide ratio from manufacturer specifications, L/D POH. L/D 0 is used whenever
available. However, the Performance Estimator requires 10 s to compute a stable L/D esti-
mate. Hence, a measured glide ratio may not be available by the time the Range Calculator
executes on the first run of the Landing Site Selector. The Range Calculator ignores the

143
current aircraft heading and calculates a perfectly circular footprint, but this is deemed ac-
ceptable since the included margin exceeds the error made in ignoring the heading by an
order of magnitude.

A.1.3 Landing Site Prioritizer

Inputs

• criteria level, c: criteria level under which landing site search is performed (1-4)

• state.loc: latitude and longitude of current aircraft position

• Engine status: engine status from the Engine Status Detector, either ”Operational”,
”Out” or ”Indeterminate”

• L rwy: sesired runway length for case engine ”Operational” to search for a runway
that fulfills level 1 criteria

• range, R: reachable range as returned by Range Calculator

• rwy info: information on location, heading, length and displaced threshold of the
selected runway

• emergency landing site info: information on location and heading of a selected

emergency landing site

• aero charts: color image files of aeronautical charts for automatic detection of un-
populated areas

• METARS: aviation weather information from built-in XM-satellite radio containing

information on wind speed in knots and wind direction in true north for the candidate
landing sites

144
Outputs

• LS[].loc: array of latitude and longitude of the runway threshold or location of the
earliest desired touchdown point for alternate landing site

• LS[].hdg: runway heading, i.e. approach direction, with respect to true north

• LS[].elevation: elevation of the touchdown zone of the selected landing site above
mean sea level

Trigger

called by higher level component, Landing Site Selector

Description

The landing site prioritizer is used by the landing site selection process to compile a priori-
tized list of landing sites meeting the criteria specified by input criteria level. The landing
site selection process is outlined in chapter 2.1.3. Figure A-2 illustrates how the landing site
prioritizer compiles an ordered list of candidate landing sites for criterial level 1, which is
the entry point for landing site selection when the engine is operational. Figure A-3 shows
the flow of functions of the landing site prioritizer for criteria levels 2-4. Criteria level 2 is
the entry point for landing site selection in case engine out, whereas criteria level 4 is used
for engine status indeterminate.
If the engine is operational, the achievable range can easily amount to one to two thousand
miles. Hence, not all runways within range should be retrieved and compared to find the
very best. Rather, autoland will search for a ”good enough” runway within increasing search
radii and stop the search process as soon as a runway is found that meets the ”good enough”
criteria. This ensures that runways closest to the aircraft’s current position are prioritized.
This is illustrated in Figure A-1.
The hard stop set by the Range Limit would realistically not be reached within the conti-

145
 Range Limit
Increasing search radii

Figure A-1: Runway search in engine operational case

nental US. In the event that no runway fulfiing the ”good enough” criteria could be found
within the range limit, autoland relaxes the criteria level to level 2, ”any runway” and se-
lects the longest runway from the set of runways within range. The good enough criteria
applied here is runway length exceeding 5,000 ft. Once at least one runway has been found
that meets this criteria, the search process is stopped. If more than one runway is found,
runways are prioritized by length. The approach direction will be chosen in such a way, that
the aircraft experiences a headwind on approach. The search process is illustrated in Figure
A-2.

In case engine out, the landing site selector initializes the criteria level to level 2, ”any
runway”. At this level, the Landing Site Prioritizer attempts to find a reachable runway
using the runway database of the FMS. If there is more than one solution, the Landing Site
Prioritizer orders the candidate runways by length. It would have been possible to include a
multitude of additional criteria such as wind conditions, runway surface, lighting, availability
of emergency equipment at airport, etc. in the prioritization scheme, but since only a small
number of runways is likely to be returned in the engine out case and runway length is the
overarching criteria for probability of a successful landing, only runway length is included

146
 Criteria level 1:
Any runway > 5,000 ft

Initialize search radius r

(r<reachable Range R)

no

Identify runways with

r>R? yes
length > L_rwy

Solution? no Increase r

yes

Prioritize runways
by length

Select headwind
end of runway

prioritized list of landing sites empty list

Figure A-2: Runway search algorithm in engine operational case

in the prioritization scheme. However, if both ends of the longest runway within range can
be reached, autoland will select the end of the runway that results in a headwind on ap-
proach. This approach direction can be deduced from XM weather information (METAR)
received at regular intervals from satellite data and runway orientation contained in the run-
way database. If no runway can be found, the landing site selection process will relax the
criteria level to level 3 and request the Landing Site Prioritizer to attempt to find a known
emergency landing site from the emergency landing site database. Since there is generally
no information on the size of the emergency landing site, the landing site prioritizer orders
the emergency landing sites within range by distance to go, in the event that more than one
solution was found. The emergency landing site database is not included by default with
the FMS, but databases, such as the ”Worldwide Soaring Turnpoint Exchange” maintained

147
 Criteria level 2:
Runway Any runway
database
Use “Nearest” function
to collect set of closet
runways
Known emergency
Criteria level 3: landing sites
Identify reachable Any known emergency
Range R
runways landing site

no Identify reachable Area charts image files

Solution? emergency landing
sites Criteria level 4:
yes Any unpopulated area
no Identify reachable
Prioritize by Solution?
unpopulated area
length
yes
no
METAR Solution?
(weather Select headwind Prioritize by
end of runway distance to go yes
information)
Prioritize by size of
area

prioritized list empty prioritized list

empty prioritized list of empty
of runways list emergency off-field landing
list list
landing sites sites

Figure A-3: Landing site selection process for case Engine Out

by the glider community, may be uploaded and stored permanently within the existing FMS
[11]. If the Landing Site Prioritizer also cannot find an emergency landing site within range,
it lowers the criteria level to 4 and attempts to find an unpopulated area from images of
the area maps. Unpopulated regions are depicted with a light green color on aviation maps,
which can automatically be searched for using color detection algorithms. MATLAB’s Image
Processing Toolbox contains routines that can be used for this purpose [20]. For landings in
unpopulated areas, distance to go is used as the ordering criteria with the approach direction
chosen as the current heading of the aircraft irrespective of wind conditions.

148
A.1.4 Initial Point Calculator

Inputs

• LS.loc: latitude and longitude of the runway threshold or location of the earliest
desired touchdown point for alternate landing site

• LS.hdg: runway heading, i.e. approach direction, with respect to true north

• state.loc: latitude and longitude of current aircraft position

Outputs

• LS.dir: direction of the traffic pattern, either right or left

• LS.IP: location of the initial point in local North-East coordinates

Trigger

called by higher level component, Landing Site Selector

Description

Given the current aircraft position, state.loc, the location of the runway threshold, LS.loc,
and the runway heading, LS.hdg, this function first decides whether the traffic pattern is
to be flown right or left. This can be deduced from the cross-product of the vector from
the current position to the runway threshold position and the runway heading vector. A
positive sign of the result indicates the pattern should be flown left, a negative sign means
the pattern should be flown right. This can be verified using the right-hand rule on vectors
V1 and V2 in Figure A-4.

149
 V2 Position of
runway threshold

V1

Aircraft position

Figure A-4: Pattern direction from cross product of V1 and V2


 −ve for ”Left” = 1
V1×V2=
 +ve for ”Right” = 0

Since the initial point, LS.IP, is given in the local North-East coordinate frame, there is
only two options for the location of the initial point:

 (0, -5,000 ft), if LS.dir = ”Left”
LS.IP(North, East) =
 (0, 5,000 ft), if LS.dir = ”Right”

Refer to chapter 2.2.2 for a description of the local coordinate system.

A.1.5 Touchdown Energy Calculator

Inputs

• LS.elevation, htd : elevation of the touchdown zone of the selected landing site above
mean sea level

• Va,td : desired airspeed on touchdown, design constant

• W: aircraft weight, design constant

Outputs

• LS.E td, Etd /W : desired energy on touchdown normalized by weight, design constant

150
Trigger

called by higher level component, Landing Site Selector

Description

The touchdown energy can simply be deduced from the elevation of the touchdown zone and
the desired airspeed on touchdown.

2
Va,td
Etd /W = htd + (A.2)
2g

with g = 9.81m/s2 .

151
A.2 Modules of Layer Trajectory Planning

A.2.1 Energy Calculator

Inputs

• state.h, h: current altitude of the aircraft above mean sea level

• state.V a, Va : current airspeed of the aircraft

Outputs

• E/W: current aircraft energy normalized by weight

Trigger

Sampling clock, every 1 s

Description

The aircraft energy can be found from the sum of potential and kinetic energy.

Va2
E/W = h + (A.3)
2g

A.2.2 L/D 0 estimator

Inputs

• state.loc, P: latitude and longitude of the current aircraft position

• state.V, Va : current airspeed of the aircraft

152
 • state.course, χ: angle of the current aircraft track with respect to true north

• state.wind: magnitude, Vw , and direction, χw , of the wind

• state.hdg, ψ: current aircraft heading with respect to true north

• state.vspeed, Vsink : aircraft vertical speed in m/s

• state.bank, φ: current bank angle of the aircraft

• E/W, Ei /W : current aircraft energy normalized by weight

Outputs

• L/D0 : aircraft no-wind, wings-level glide ratio

Trigger

called by higher level function, Performance Estimator, executes every 1 s

Description

This module measures an instantaneous, apparent glide ratio L/Di and corrects this value
for bank angle and wind effects to arrive at the no-wind, wings-level glide ratio, L/D0,i .
L/Di is calculated from the change in energy, ∆E/W , and distance travelled over ground in
a given time interval, ∆R. After correction for wind and bank angles effects, the samples of
L/D0,i are passed through simple moving average filter with a sample depth of 10 to arrive
at smoothed measurement of the no-wind, wings-level glide ratio L/D0 . With a sampling
time of 1 s, the first L/D0 is therefore available after 10 s of the initiation of the Performance
Estimator. The change in energy, ∆E/W , over one sampling interval becomes:

∆E/W = Ei /W − Ei−1 /W (A.4)

153
The distance, ∆R, travelled within one sampling interval becomes:

�
∆R = (Pnorth,i − Pnorth,i−1 )2 + (Peast,i − Peast,i−1 )2 (A.5)

The instantaneous, apparent glide ratio, L/Di can be written as:

∆R
L/Di = (A.6)
∆E/W

This L/Di is now corrected for wind and any non-zero bank angle to arrive at the wings-level,
no wind glide ratio L/D0,i . A non-zero wind has the effect of changing the distance travelled
over ground when compared to a no-wind situation. Hence, it needs to be removed to arrive
at a no-wind glide ratio. The approach taken here is to break up the wind into a crosswind
and tailwind component and remove their effect on L/Di separately.
Note: Vertical wind speeds strongly affect the glide ratio. However, since this wind cannot
be inferred by the wind estimator, its effect cannot be removed. Therefore, vertical wind
effects are absorbed in L/D0,i .
Va
Tailwind increases the no-wind, wings-level glide ratio from L/D0,i ≈ Vsink
in a no-wind
situation to L/Dw,t ≈ L/D0,i + VVsink
w,t
in case of a perfect tailwind. The magnitude of tailwind
Vw,t can be deduced from the course angle, χ, (angle between the inertial north axis and the
ground speed vector), the wind direction, χw , and the magnitude of the wind speed returned
by the wind estimator as illustrated in Figure A-5. The equation for the tailwind component
can be written as:
Vw,t = Vw · cos(χw − χ) (A.7)

If χ and χw differ by 180◦ , Vw,t is a perfect headwind denoted by a negative sign. Non-zero
crosswind results in a non-zero crab angle, χc . This non-zero crab angle causes the distance
flown over ground within a unit of time to be larger than the distance flown in the airmass
within the same time interval. From Figure A-6, the crab angle, χc , is the difference between
the aircraft course, χ, and the aircraft heading, ψ, i.e. χc = χ − ψ. Hence, L/Di uncorrected
for crosswind effects will be larger than L/D0,i . The correction factor becomes cos(χc ) since

154
 )= ∆(E/W r m
�� � 1�
16000 16� n i=1 Ri nE
R161R =n+ �
� (51) av
�
L/D_2 rr =
∆R )= (∆pos.east) + (∆pos.north) 16000
2 =+2m(∆pos.east)
(L/D)
∆∆R(L/D) Rn
(∆pos.north)
n2 nn n
Figure 14: Aircraft glide ratio estimation
(46) 2 2
s(ψ)
= const L/D =
(L/D) L/DL/D
i=1 av
ro =av==
n 16000
−
av =�n av
i i=1
ri=1
∆R rR=
1316
ii
+
i=1
R
i
= �ni=1 i=1 i (49)
n
E(L/D)
(8) 2 − E1 (49)
∆R (49)
i (45)(46)(47) (6)�(52) n (
= nn 16000 n16
av
Aircraft glide ratio estimation 4L/D 13000 av�
i=1 Ri L/D = (51)
E2 − E1 L/D_3
L/D_n
L/D = �
(52) R ∆R
∆(E/W
av
L/D
Then = parameters
oldest
� � n R)n iL/D
av
=
sample out
i=1 n
from (L/D)=
the ∆Rn
W
databus
i L/D for
= (51)
∆R
determination of (47)
apparentE avglide
= i=1
rat
W
ermination of apparent glide ratio R
E av
av =
E E
mavavRn i=1
∆(E/W
E == )
R1 av L/D
i=1 n
i=1
R iEi
= � n n (50)h∆(E/W (50)
∆(E/W
(50) )
)(47) (48)
n
(47) L/D
are: r = WW i=1 � n n �
•
av+ Airspeed, i V and altitude for determination (46)of energy
� � �
W r
R2 av Figure =E 16000
av
14: Aircraft
− E n a
16 nglide ration estimation ∆(E/W
i=1 Ei 4 E)2 − E1 �n a
(
V (L/D)
nation
=H ∆Rof= 0 · cos(ψ)
2
(∆pos.east)
energy
(∆pos.east) 2 2
+ (∆pos.north)
+ (∆pos.north)
4 ∆R = (∆pos.east)2 + (∆pos.north)2
E 22
�
−
E
�
E 1
• n 1
GPS �
=RχE and
�ni=1 E
n avi =
positionE in
i=1 E E i
North-East
av
(9) coordinates (52) at sample L/Dav = i=1
intervals
(52)
(48)
th axis and the ground speed vector),
The parameters from the thedatabus
wind direction W
for
L/D
� W
n
avNorth
n= =iw av
Edetermination i=1i=1
n =E(L/D)
i
the
of Eapparenti
magnitude
av glidenofratio are: W (48)(49) (7) n(
nni=1 Ei 2πr nWE (51) − E
i=1av
av
sampleoldest sample outL/D =in Figure R = (47)
V sbyatthe
atio
intervals
Vare:wind estimator as illustrated
• Airspeed, Va and −→altitude
E
∆R
−→
L/D
av ??. ∆R
=
h for = 5.2 −−−∆R →
)+=determination E�
Correction
in+
n n
of energy
W E � of −−→ 2 apparent
the E −→ 1
(51)
−−−→glide ratio
4
2πrfor wind Eand
� R =|
4 • GPS position in North-East
P I | +∆(E/W
E | av
F B I
L/D |
= | T
∆(E/W
�
d F i=1
B | −2d
)
n
�
E
n
π− (L/D)
| �n must now
∆crs |
(53)
(51)
R2=|nP I | + |1FB I | + | Td FB | −2d + (48)
av
� forNorthwind
(∆pos.east) 2 +and (∆pos.north)
av ∆(E/W
coordinates n 2 )
at Vg
nsmoothed
av (L/D)
i=1 intervals
sample
i=1 i i
(L/D) | ∆crs |
π−(50) W
glide ratio
R 7:= Aircraft
re (∆pos.east)
glide ratioV
�
2 +estimation
bank
E2 − E1 �angle
(∆pos.north) E
L/D
E =2
L/D
av1 − cos(π−
−
�
TheEav
2W2 1L/D
av
i=1
1 2 ==E
effects
L/D
|4
i W
∆crs |)
n
avn L/D’s
4= i=1 (L/D)
Wi=1
� W be� � �
i = (∆pos.east)
(52)
∆R
i
corrected
(52)
for any(48)
(49)
wind and
2 + (∆pos.north)2
(49)non-zero
4 ban

stfor
=any ∆R5.2
wind =and+ Correction
(∆pos.east)
non-zero
4 2+
w,t
bank
W
of(∆pos.north)
the
d = rapparent
angle
E
to arrive� − E
level,
n at
n
nglide no
the
wind
avratio
(L/D) = component
wings-
tailwind
glide
for wind ratio
∆R (L/D)
and n ==bank
.
0 1− Figure
angle
(54)
(52)
15
(∆pos.east) illustrates
| ∆crs |)
cos(π−effects
2 + (∆pos.north)
2
how +(L/D)
(8) observe(
(∆pos2
(L/D) 0
L/D =
1 + cos(π−
� tailwind
W i=1 E
Eav
|(L/D)
∆crs |)
component ∆R i and n= (10)d(∆pos.east)
crosswind r component.
1 + cos(π− | The
∆crs amount
|) of tailwind can
4 the ∆R
15 illustrates
d bank angle how
Theeffects −→V
(L/D)
Vg ·− sin(γ
→ −I→L/D is− −−
av
−
0→ →)
corrected − −−→
n
for 2πr
i=1(L/D) bank av angle,i
2πr (50) ∆R
e corrected R
forsmoothed
=| P
any I | a
+ L/D’s
observed
wind|
windR glide
F B
=| can |must
+
and
| be
P Iratio |
+L/D−
T now
→ F |
non-zero
Fav
d
|(L/D) B
B Iav
=
be−2d
| 0+=
corrected
+ angle
| Figure
Td Fπ− Bbank
i=1 || 15
forWanyi wind
χ
−2d −
(angle
n
L/D
+ →
angle ∆R and non-zero
to=∆R
between arrive bank
at −
(53)
inertialtheangle to arrive at the wings- (50)(49)
− −
north
→ axis and
(53) for bank the
(49)
L/D_n ground speed
L/D
2π
vector),
=
+− |→ n− → −(L/D)
−
for− → 2πr
nt. The amount �of notailwind calculated from ∆crs the | course
ank angle to1
level,
arrive at(∆pos.east)
the wings-
tailwind
−→ observed
(L/D) .
−→ = (L/D) L/D n =
L/D
−−−→windT,windC,bankW
illustrates
π−= 2πr
| ∆crshow (L/D)
| speed
E observed is corrected
(55) angle,
(51) ∆(E/
D)
γ)
sved 0 .=
and Figure
is the
GR
∆R =??
� tailwind
ground
corrected
∆Rcomponent.
bank R
illustrates
for speed
0
= (∆pos.east) 2R
component =|
vector),
angle,
=+
�+
2
component
R =|
=|
and how
1 −the
r (∆pos.north)
P
|crosswind
P (∆pos.north)
Ithe
Pwind
cos(π− I
(L/D)
F�
| ∆crs |
B I |direction
|2hence
I�tailwind +
|2magnitude
d FB | −2d
+component.
Tobserved
+|) axis|becomes:|F
χwThe
andB
F isofamount
+∆(E/W
and
IB
π−
Vw
the
the EI
corrected
| |∆crs
of|
)wind
+|speed
∆(E/W
av +
∆(E/W
av |
| Tcrosswind
tailwind
dF )can T
returned
bank
observed
Bthed
)F by
=(53)
|wind−2d
be calculated (L/D)
|
thefrom
wind
−2d course +The tailwind c
the estimator.
windT,windC,bank
B +χw and the L/
1 (
wind
anwind
),
be
the wind
angle χ
estimator.
calculated fromThe
(angle
theThe dbetween
tailwind
course amount 1
component
+
magnitude Tof the wind speed returned
direction χ and the
d =
inertial
cos(π−
ofr
1 −north
|
(L/D)∆crs E
cos(π−
E Eavav
|)
by
av
| component
∆crs
E 2 −
the
|)E groundand
∆R
the wind estimator. The
windT,windC
L/D =
1
W W
vector),
tailwind
East
(54)
component(56)
direction
(54)
hence Vw,t = π−
(50)
becomes:
(51)
−→ →
dot(Vw|, V
(52)
−
c ) π−
∆crs E2 − E
| | ∆
ector Vw = −→[x, − y, 0] Vwprovided by the
→ 1 + cos(π−
1 −autopilot
W | ∆crs
cos(π− | ∆crs and|)
|) the current− E(55) course (L/D) (50) W
2w ,E
w

)
Vdw,t �
cos(ψ)
� ·
= dot(V
component �
cos(ψ)
w , Vcbecomes:
hence )
(L/D) =
d=r
(L/D) W Wdirect
A V| w,t
W
∆(E/W
�
headwind
= |)dot(∆R E� −→) −
V(8)
→
− 2
)E1 yield
Vcwould 1 a negative
windT,windC
(54) Vw,toldest sample out ◦
due to the (9)
(8) 180 difference in V
= 0 (∆pos.east) 2 2 East 2 1 + cos(π−
L/D = ∆crs
2 (51)
4 (8)+ (∆pos.north)
= (∆pos.east) + observed
(∆pos.north)L/D_n windT,windC,bank
∆RVto = (∆pos.east) 2 2
2+ (∆pos.north)
Figure (L/D)
A-5: − →
Inferring = (L/D) − → − − − → 2πr (55)
. |tailwind component ) ◦from ground speedV and
(L/D) (57) wind speed
,t2 due V
Vthe A 180 ◦
difference 0]Tin . V2R
observed
w =| and IaV|negative
Ptailwind
c+ FB windT
|E+2due
IVcomponent |∆(E/W
windT,windC,bank
− TE F W
the| −2d + W (53)
)ector
n cos(φ)
5:
c 2=3/2
VInferring
+[cosχ,
+ (∆pos.north)
s.east)
direct headwind
sinχ,
(∆pos.north)
tailwind component (L/D)
would
from
The
(L/D)
yield
ground = (L/D)
observedspeed − cos(π−
1 −1cos(π−
and∆R | ∆crs| |)∆crs |)
wind
w,t
W speed
windT,windC,bank
dto1B 180
hence difference
(56)
π−becomes: in Vw and(L/D)
| (11)
∆crs
(55)
Figure
| 14:
17
c . windT
Aircraft glide(52) ratio estimation 4

→ dd− = r r −−−∆R
w and Vc . �
= const V17w,t =Vdot( a = V
−→ → −
Vgw· ,cos(χ
Vc ) −4ψ). − −→
−
L/D
The =
windT,windC
(L/D)
→−→
→ effect
=(L/D)
→−−−
∆(E/W →
∆R
windT,windC E 2
of 1non-zero
0 −−
The
−
2πr
)− E
2πr
1
−→ bank
cos(π− 17
parameters | ∆crsangle
from|) the is(12)
to reduce
databus (1)forthe
(56)
(58)
wings-level,
determination
(51)
of no-bank
apparent
(54) (52)
glide ratio are:
tailwind component hence becomes:
oldest ◦sample RR=|R =|
(L/D)out
L/D = L/D
P I
windT
∆R
=|PPI || ++| |FFI
| |I
++
L/D
L/D
(L/D)
=
1
d| =
| |
= =
−→
+
T
F+1
|
FBT+
cos(π−
rI |∆R
| −2d
F +
1|
+ +|
cos(π−
−2d
|∆crs
T
W+ d|F|)
cos(π−
•BAirspeed,
B ∆crs
B | |)
−2d
| +
∆crs(
Va dandπ−
|)
2πr
Baltitude
d
| ∆crs B(57)
(L/D)0
(56)
h| for determination
(53)
(51)
ive V due
= (L/D)0 ·Vcos(φ) to the 180 3/2 difference (L/D) ∆(E/W
in
(L/D) 2V)
∆(E/W−
�
∆(E/W π− | ∆crs
windT,windC
EwindT,windC,bank
Eand
) 1 π−| | ∆crs
V ) . −
→ | −→ −→ (57)
of energy
w,t = V · ∆(E/W )
cos(χ − χ)
windT
w North c
R =| P I | +
R |=|
F IP (13)
(8) (59)
(L/D)windT,windC,bank (52)
w,t
worked out similarly using the
w
��
w
(L/D)
(L/D)(cosφ)
cross
0 d =
W 3/2
product
observed
r
1− •=GPS (L/D)
byposition
cos(π− | ∆crsin|) North-East
extracting
windT,windC,bank the(58) coordinates
third 3/2 (54)
(55)
at sample intervals B
(L/D) (57)
(cosφ)
◦
0 , Vw,t Figure
esults in→
is a perfect
1
−a non-zero V (L/D)
7: Aircraft
headwind−
V(L/D)
crab −
R→ E d−
=|→
angle
glide
→denoted
−P(L/D) E
ratio
1−
by
− −=
→aestimation
−
(L/D)
E
I c| if+windT,windC,bank
=
χ r |F
the I |
aircraft E+ (L/D)
negative
E1cos(π− =
(L/D)
Ecos(π−
−
2 | −T
windT
|E∆crs
follows | |)∆crs |)
(L/D)
−−− E
5.2
F
w,t cosψ(L/D)ifwindT,windC
2
1 + cos(π−
0sign.
→
observed
windT,windC
observed
1 |
4 | ∆crs
Magnitude
Correction
−2d
course1 χ+ as 2
of Va
distance flown per
shown
no wind present2
2πr
�(
|)
defines
unit of time
1of
in the
Figure 1 apparent
(58)
windT,windC,bank
windT
glide
(60) ratio
(53)
4
,windC,bank
for
(56)wind(52) and bank
R
L/D) 0 ·causes d = r w,t B d B
π−be| larger
∆crs |than (14) (L/D)windT,windC
D) + + V · sin(γW ) 1 +1cos(π−
angle
must0
V =thecross(
(cosφ)
now be
distance
1.5
0corrected
Vwflown
Vw,c
, Vc )over ground
for any(cosφ) wind(L/D)
3/2
(L/D)
interval. Hence,�(L/D)observed
within
+ cos(π−
W
and | ∆crs| |)∆crs |)
W
(L/D)
observed
W
a unit
non-zero
The = of
(L/D)
0 smoothed
windT,windC,bank
time
bank (10)
to
windT,windC,bank
Vw L/D’sangle mustto arrive
(59)
now beat
(2)
the(58)
corrected (55)
for any wind and non-zero
cosψ 4 bank angle
atio V · sin(γ )
airmasswith
(L/D)
within
n (L/D)0 . The
V .
0w,c
the same
ψ = asin(
=
Figure V
correction 8
)
Va(3) a
illustrates
afactor(L/D) is d cosχ
=
how 1 −
(L/D) 0
cos(π−
(cosφ)
cr since Vtas with
windT,windC
level,
3/2
0
Vauncorrected for crosswind
|
Vg(L/D)
=observed
ψ
no
cosχ
=
∆crs wind
is |) d=r
glide
corrected
c .windT
asin(
Vw,c
)
ratiofor (L/D)
bank 0 (3)
. Figure
angle,
(59)
15 illustrates
(54)
how
(57) (L/D) observed is co
sswind • Airspeed, V a−→ − → −− −→V|w,t
tailwind
(L/D) component and 2πrcomponent
crosswind
(60) component. The amount
(56) of tailwind can be calc
ed from component.
−→
Airspeed,
atabus the
for −→ (L/D)
databus
V correctionforThe
→ −−−
−→
(L/D) amount
correction
of the −→→ of of
thetailwind
(L/D)
= (L/D)
cosψ
(L/D) −−−(L/D)
1+
(L/D)
(L/D)
component
→−−−→ 2πr 2πr
cos(π−
measurement
windT,windC,bank
measurement ∆crs and
|) a crosswind
are:
windT,windC
are:
V
1
1 V•wWind R a
=| P|−I | + | F− → (L/D)
I | +
(L/D)
= | T F −
angle
observed
windT,windC
windT χ| −2d
2πr Vg
(angle +
between the inertial
windT,windC,bank north (61)
(59)
axis and the ground speed (53)
vector), the win
R =| PP II ||++| F| FI | I+(L/D)
T
R =| P I | + | F RI=| + | T F | −2d
byB the autopilot
T F|+T| −2d dobserved
3/2B (60)
|| + + (
ector V= [x,vector,
y, 0] Vprovided (cosφ) and
Vsink the current windT,windCcourse windT ,windC,bank Vw,t
B
F
cosψ
| d
−2d +B B dπ− B | ∆crs |
d BwindT ,windC
h ψ = asin( w,c
) w magnitude (L/D) of the wind speed returned by the
(L/D)windT wind
− estimator. (58) The tailwind compon
dWind
GR by unit vector,
0 Vvector
a
VVw = [cosχ, (L/D) sinχ, 0]T . The π− | ∆crs
= (L/D) π− | ∆crs
tailwind
(L/D) | | B 0
component hence becomes: Vsink(55)(57)
• Current
with ψ = asin( Va )
c
Vw,c
course
� flown χ 4
observed
� (L/D)
Vw,t windT,windC π− | ∆crs(L/D)
windT,windC,bank
windT
| (
East
V = dot(
−
→
V
→
−
w , Vc )
observed
Current course10flown χ (L/D)
• Bank angle,
−
φ = dot(Vw , Vc ) −
Vw,t
�
→windT
(L/D)→
− 1 −
�| ∆crs |)Vsinkcos(π−
(L/D)
cosψ | ∆crs
windT,windC,bank (61)
|)would yield a negative
(60) w,t
(1) Vw,t due to the 180 (59)
1 − cos(π−(L/D) V(L/D) 3/2windT,windC ◦
wn
Bank χ angle, φ Figure d = A-6: r (L/D)
Effect
1 − cos(π− | ∆crs |)
(L/D) 4
A direct
ofwindT,windC
wind − on
(cosφ)
w,t headwind
horizontal
0 distance
windT travelled
,windC per unit of (56)
(61) time(58) (54)
difference in Vw and
• with d=r
Sinkψrate,= asin(
Vw,c
Va )
Vsink d = r ◦ 1 + cos(π−
1(L/D)
windT
− cos(π− Vsink| ∆crs |)
| ∆crs |) ( 4
windT
ield
Sinka 4rate,
negative
nt can be worked glide out
1 + cos(π−
ratiosimilarly
d = r 1|+∆crs
Vv Vw,t due to the 180 difference(L/D)
using the
(L/D)
|)
(L/D)
cos(π− | ∆crs |)
φ)3/2
(L/D)
cross
in Vw
Vproduct
and Vc .
windT,windC
windT,windC,bank
(L
by extracting the and windT
thirdbank angle (57)
17 (60)
V (L/D) (L/D)
=
to L/D0,i
(L/D)
windT − (cosφ)
V
· (cos
3/2 1 +windT
(L/D) cos(π− | ∆crs
.windT
(61) |)
Including
w,tcosψ tailwind,
(55)
crosswind effect,
(59) yields
the=→
ψ
V
−following
with (L/D) observed
−
→ →
asin( V )formula
observed
− 4for= (L/D)
L/D 0,i
w,c
from (L/D) (L/D)
windT,windC,bank
sink
i . windT
windT ,windC,bank (2) (11)
3/2 V = cross( Vw , V c ) a
4 windT,windC
φ) (L/D)
(L/D) (L/D)
observed
0 = (L/D)
Vw,t windT,windC,bank
(58)
(60) (
/2 Vw,c = V (3) (L/D)observed
cosψ
(L/D)windTV− · cos(χ − (L/D)
= (L/D)
(L/D) 0 (3)− ψ) 4
χw ) windT,windC,bank
cos(χ (61)
st with ψ = asin(
Vw,c
) L/D = [(L/D) (L/D)
i−
windT,windC
4 0,iwindT,windC,bank
(L/D)
w Vsink
Vare:
0
(L/D)
]· 3/20 (L/D) (A.8) (56)
(12)
from Figure 16:
the databus forAircraft wings-level,
correction
Va of the no-wind
(L/D) glide ratio relation
measurement
(L/D)(cosφ)
windT
sink
3/2 ,windC 0 (cos φ) (59)

(L/D) −
4 (L/D)VwindT,windC
w,t (
(L/D) in glide ratio. A 20 kt(61)
parent glide
D) · cos(φ)
0
Theratio
tailwind
11
component
3/2
estimator has
(L/D)
from
(L/D) the L/D
windT effect
largest
windT,windC
windT (L/D)
0, on the and
wind,
Vsink
windT,windC,bank (L/D)
changenominal
windT,windC,bank
windT,windC (60)
tailwind
(57)
(13)
cosψ by(L/D)
k angle increases the 3/2
· cos(φ) observed (L/D)
glide ratio windT 22% for(cosφ)
windT
an 3/2 ,windC,bank
airspeed
windT (L/D)
of 90 kt. A 20 kt crosswind for a 90
3/2 ,windC,bank
0 Vw,c
h ψ = asin( Vakt) airspeed creates a 13◦ crab angle, which4
(L/D) (cosφ)
increases
windT the 3/2
an apparent
estimate ofglide
(cosφ)
the ratio by only 3%. A
(L/D(
calculate the lengths of the legs of the traffic pattern, autoland
(L/D) 0 (L/D)
requires 3/2 (cosφ) (58)
ide ratio (L/D)expected on each leg of the traffic pattern. Using the(L/D)
windT,windC
(L/D)0 returned by the L/D
1 30◦ bank angle as nominally held during
(L/D)
(L/D)expected can be calculated using knowledge (L/D)
of the−
V w,t the
nominal
(L/D)4 turn segments
bank
windT decreases the glide ratio by 20%.
angle during turns and the (61)
1and tailwind experienced on each leg (from
windT
wind estimateV0sink
and cosψ
knowledge
After having applied the corrections for wind and bank
1.5
windT
of the ,windC
windT,windC
course of(L/D)
each
angle effects, the sampled L/D0,i are (14)
cosφ)
rocess of calculating (L/D)expected(L/D)
Vw,c
then becomes (L/D)
the inverse process
windT,windC,bank cosψ windT
of the
0 process
(L/D)
illustrated in
,windC 4(
(L
ith ψ1.5
φ) = asin( Va ) (L/D) (cosφ)3/24 4 155(L/D)
windT ,windC,bank 0
cosψ (59)
Vw,c
= asin(
nd estimation Vw,c
Va ) (cosφ)(L/D) 3/2
Vw,t Vw,c
cosψ
asin( ) (L/D) windT,windC,bank
with ψ = asin( Va )
windT,windC
(L/D) −
17
windT
Va and course of the wind is returned by the autopilot,3/2which (60) (
e ofVthe
w,cmagnitude
(L/D) cosψ (L/D) (cosφ) V sink Vw,t
will be used to
Vw,c
( )
he crosswind and tailwind components
V
windT
needed for ,windC
(L/D)
estimation
windT,windC,bank
of the no-wind,with
windT −
wings-level ψ = asin(
aircraft
V )
averaged using n = 10 samples to arrive at the estimated aircraft glide ratio L/D0 :
�n
i=1 (L/D)0,i
L/D0 = (A.9)
n

156
A.2.3 L/D expected Generator

Inputs

• L/D0 : aircraft no-wind, wings-level glide ratio

• LS.hdg, χr : runway heading, i.e. approach direction, with respect to true north

• TP.leg[].course: array of courses of the traffic pattern legs in local North-East coor-
dinates

• V glide, Vbg : best glide airspeed as commanded to the autopilot, design constant

• state.wind: wind speed, Vw , and direction, χw , as returned by the autopilot’s wind

estimator

• nominal bank angle φ0 : nominal bank angle to be used for turning

Outputs

• TP.L/D exp:[] array of L/D expected, one per traffic pattern segment

• HP.L/D exp[]: array of L/D expected, one per holding pattern segment

Trigger

called by higher level function, Performance Estimator, executes every 1 s

Description

This module takes the L/D0 estimate returned from the L/D 0 Estimator and estimates the
expected glide ratio for each leg of the traffic pattern, TP.L/D exp[], or holding pattern,
HP.L/D exp[], using the expected crosswind and tailwind components, plus nominal bank

157
angle for turn segments. For this purpose, the L/D expected generator first calculates the
course of each traffic pattern leg in global coordinates, which by rotating the courses given
in TP.leg[].course by runway heading, χr , a function that is performed by library module
”convertToGlobalCoords”. For the holding pattern, the course of the outbound leg is known
to coincide with runway heading, χr , and the inbound leg being the opposite direction, i.e.
χr − 180◦ normalized to range ] − 180◦ , 180◦ ]. then for each traffic pattern segment and hold-
ing pattern segment, the L/D Expected Generator calculates TP.L/D exp and HP.L/D exp
respectively using the expected wind and bank angle along the segment. For turn segments,
a nominal bank angle of φ0 is assumed and zero bank angle for straight line segments. For
calculation of the crosswind and tailwind components during the turn segments, the center
course between the course at the start of the turn and the finish of the turn is used. The
crosswind manifests itself as a non-zero crab angle, χc , which can be calculated from the
wind speed, Vw and directionχw from the following equation:

Vw · sin(χ − χw )
χc = tan−1 ( ) (A.10)
Va

where the term Vw · sin(χ − χw ) denotes the magnitude of the crosswind. The tailwind
component adds to the forward speed of the aircraft and increases the glide ratio by the
Vw ·cos(χ−χw ) Va
term Vsink
, with Vsink ≈ L/D0
. The bank angle is included through term (cos φ0 )3/2 .
In summary, L/D exp can be written as:

(cos φ0 )3/2 Vw · cos(χ − χw )

L/Dexp = L/D0 · + (A.11)
cos(χc ) Vsink

In order to include the effect of wind shear in boundary layer, an ”effective” wind speed
Vw can be calculated for the traffic pattern legs, where wind shear is expected to have an
impact. For a wind profile, where the wind speed changes linearly with altitude with no
change in wind direction, the effective wind speed can be calculated from the wind speed
expected at the end of the leg, Vw,2 and the expected change in wind speed from the start of
the leg to the end of the leg, ∆Vw = Vw,1 − Vw,2 . Both Vw,1 and Vw,2 can be inferred from the

158
assumed wind profile and the altitude of the aircraft at the start and end of the given leg.
This altitude is known from the planned trajectory and the expected aircraft performance.
The effective wind, Vw becomes:
Vw = Vw,2 + ∆Vw (A.12)

A.2.4 Path planner to Initial Point

Inputs

• LS.IP: location of the initial point in local North-East coordinates

• LS.loc: latitude and longitude of the runway threshold or location of the earliest
desired touchdown point for alternate landing site

• state.loc: latitude and longitude of current aircraft position

• range, R: achievable range

• L tp,0: length of the baseline traffic pattern

• LS.hdg: runway heading, i.e. approach direction, with respect to true north

• engine status: engine status

Outputs

• LS.LS Ok: flag indicating whether the provided candidate landing site is feasible
given the range, 1=success, 0=fail

• TI.leg[].length: array containing the length of each leg of the trajectory to the initial
point

• TI.leg[].fix: array containing the waypoints that specify the trajectory to the initial
point in local North-East coordinates

159
Trigger

called by higher level function, Landing Site Selector

Description

The task of the Path Planner to the Initial Point is to generate a feasible path from the
current position, state.loc, to the initial point, LS.IP, so that the aircraft is aligned with the
course of the downwind leg by the time of arrival at the initial point. The path planning
is performed in the local North-East coordinate system. This is justified since the typical
distances to a landing site are expected to be 100 miles or less, which creates an error of
less than 0.002%, or 10 ft in total. This error is not noticeable in comparison to the margin
included in the achievable range, R. With this approach, the path planner first calculates
the aircraft’s current position in local North-East coordinates using library function ”con-
vertToLocalCoords”. The initial point, LS.IP, is already provided in local coordinates. If
there are no obstacles along the straight line path from the current position to the initial
point, the Path Planner to the Initial Point chooses the simple straight line trajectory and
inserts an additional leg that allows the aircraft to align with the downwind leg upon arrival
at the initial point. If there are obstacles along the straight line path, a segmented trajec-
tory will need to be synthesized. For this purpose, the path planner uses a digital terrain
map. If the engine status = ”Operational”, the aircraft is capable of holding altitude so
that the 3D problem reduces to a 2D problem. From this 2D map, the path planner deduces
terrain elevations greater than the aircraft’s current altitude and represents these elevations
as polygonal obstacles. In order to deduce an obstacle-free path from the aircraft’s cur-
rent position to the selected destination, a representation of feasible trajectory segments are
stored in form of a roadmap. This is most commonly done via a visibility graph. This graph
connects each node of a polygon with the node of another polygon, given that the nodes are
within line of sight, i.e. do not cross an obstacle. Figure A-7 illustrates this concept. Since
the full visibility graph contains many edges that are not part of the shortest path solution,

160
 Figure A-7: Nodes and edges connecting the nodes on a visibility graph

the full visibility graph can be reduced to a reduced visibility graph. From this point, search
algorithms, such as A∗ , can be used to find the shortest trajectory from the start to the goal
by setting the cost of each edge equal to the length of each edge. For the cases engine status
”Out” or ”Indeterminate”, the simple 2D map will not suffice. A 3D representation will
have to be generated taking the ceiling of each obstacle into account as opposed to only the
obstacles that exceed the aircraft’s current altitude. Meuleau, Plaunt and Smith propose to
use an extended tangent graph, which includes additional edges that take into account the
aircraft’s capability to overfly an obstacle [?]. The path planner passes the final trajectory,
as a list of waypoints in local coordinates via TI.leg[].fix. It stores the length of each leg in
TI.leg[].length. If the total length of the synthesized trajectory is greater than the achiev-
able range R, the path planner sets flag LS.LS Ok to 0 to indicate failure to find a feasible
trajectory. Otherwise, the path planner set the flag equal to 1.

A.2.5 Flight Plan Feasibility Monitor

Inputs

• L tp,0: total length of the baseline traffic pattern, design constant

• TI.leg[].length: array containing the length of each leg of the trajectory to the initial
point

• TI.L: distance to the next waypoint

161
 • TI.currSeg: integer identifying the current leg the aircraft is on

• range, R: achievable range as returned by the Range Calculator

Outputs

LS.FP Ok: flag denoting whether the target touchdown point can still be reached given
the current energy, E/W and the glide ratio estimate L/D 0, 1=OK, 0=NOK

Trigger

started and stopped by higher level state machine, executes every 3 s while running

Description

The Flight Plan Feasibility Monitor compares the achievable range, R, to the remaining
distance to fly until the target touchdown point, which is simply the sum of the remaining
distance to fly to the initial point plus the distance to fly in the pattern. The distance to
fly to the initial point can be calculated from knowledge of the distance remaining on the
current leg, TI.L, the length of the remaining legs stored in TI.leg[].length. The remaining
legs start at index, (TI.currSeg + 1) = k, in array TI.leg[].length. The number of elements
in array TI.leg[].length is denoted by n:

n
�
Lip = T I.L + TI.leg[k].length (A.13)
i=k

This calculation neglects the fact, that the aircraft cuts the corner between two subsequent
legs to perform a smooth intercept to the next leg. This can be neglected for small course
changes from one leg to the next (< 45◦ ) since it is expected that these errors are offset
by cross-track errors the aircraft will inevitably incur along the trajectory. Overall, this
approach overestimates the length of the actual trajectory, so that it errs on the conservative

162
side. For trajectories including many and many sharp turns, it is advisable to adjust this
scheme in a subsequent design iteration. Hence, the flag LS.FP Ok is set according to the
following conditions:

 NOK = 0 if R < L + L
ip tp,0
F P OK =
 OK = 1 if R ≥ L + L
ip tp,0

A.2.6 Excess Energy Calculator

Inputs

• E/W: current aircraft energy normalized by weight

• HP.R, R: range to go until start of the next segment of the holding pattern

• HP.curSeg: integer specifying the segment of the holding pattern, the aircraft is
currently on (1=turn to outbound leg, 2 = outbound leg, 3 = turn to inbound leg, 4
= inbound leg)

• LS.E td, Etd /W : desired aircraft energy on touchdown, design constant

• r: nominal turn radius, design constant

• L: length of a single leg of the traffic pattern.

• HP.L/Dexp[]: expected glide ratio on each segment of the holding pattern

• HP.length, L: length of the current inbound and outbound leg

• TP.L/D exp[]: expected glide ratio on each segment of the traffic pattern

• L: length of a single leg of the traffic pattern

• L m: runway distance consumed by undershoot margin

163
Outputs

• HP.dE/W, ∆E/W : difference in energy between the estimated aircraft energy upon
passing the initial point next, and the energy required to start the baseline traffic
pattern adjusted for the current aircraft performance

Trigger

started and stopped by higher level state machine, executes every 1 s; paused during turn
segments

Description

The Excess Energy Calculator estimates the excess energy, above the energy required to
start the traffic pattern, upon completion of the current loop. This excess energy is used to
determine the length of the inbound leg and outbound leg of the holding pattern, and the
number of loops to perform in the hold after completion of the current loop. For this purpose,
it first computes the energy required to complete the current loop, Eip /W . Calculation of
the excess energy is halted during the turn segments. Hence, there are only two cases to
consider: 1) The aircraft is on the inbound leg or 2) the aircraft is on the outbound leg.
It would have been possible to schedule only one calculation of the excess energy estimate
by the time the aircraft starts the inbound leg. However, since there is the possibility of a
zero length inbound leg, which causes the holding pattern update to arrive too late for a
smooth transition onto the downwind leg should the update have commanded an exit from
the holding pattern.
If the aircraft is on the inbound leg, Eip /W can simply be calculated from:

Eip /W = R · (L/D)exp,4 (A.14)

164
where L/Dexp,4 represents the estimated glide ratio on the inbound leg. If the aircraft is on
the outbound leg, Eip /W can simply be calculated by adding the turn from the outbound
to the inbound leg and the inbound leg:

Eip /W = R · (L/D)exp,2 + rπ · (L/D)exp,3 + L · (L/D)exp,4 (A.15)

where L/Dexp,2 represents the estimated glide ratio on the outbound leg and L/Dexp,3 , the
expected glide ratio during the turn onto the inbound leg. The excess energy ∆E/W hence
becomes:
∆E/W = E/W − Eip /W − Etp /W − Etd /W (A.16)

with

L−r rπ L − 2r rπ L − r − Lm
Etp /W = + + + + (A.17)
L/Dexp,1 2 · L/Dexp,2 L/Dexp,3 2 · L/Dexp,4 L/Dexp,5

where Etp /W is the energy required to fly the baseline traffic pattern with the current
expected glide performance.

A.2.7 Holding Pattern Generator/Updator

Inputs

• HP.dE/W, ∆E/W : difference in energy between the estimated aircraft energy upon
passing the initial point next, and the energy required to start the baseline traffic
pattern adjusted for the current aircraft performance

• HP.L/Dexp[]: expected glide ratio on each segment of the holding pattern

• r: nominal turn radius, design constant

• φ0 : nominal bank angle, design constant

• L/D 0: estimated no-wind, wings-level glide ratio

165
Outputs

• HP.#loops, n: number of loops to spend in holding pattern after completion of the

current loop

• HP.length, L: length of the straight line segments of the inbound and outbound leg
of the holding pattern

Trigger

started and stopped by higher level state machine, executes every 1 s; paused during turn
segments

Description

The Holding Pattern Generator/Updator recalculates the number of holding pattern loops
to follow, n, and length of the straight line segments, L, every time it receives an update on
the energy that is to be depleted in the holding pattern after completion of the current loop,
HP.dE/W. The current loop of the holding pattern is not altered, since the holding pattern
generator cannot update the length of the current hold loop, while the FMS is tracking it.
The integer number of loops, n, that can be completed with the excess energy, ∆E/W , can
be found from:
∆E/W
n=� � (A.18)
Eloop /W
where
2rπ
Eloop /W = (A.19)
(cos φ0 )3/2 L/D0
and �x� represents rounding down. The length of the straight line segments, L, becomes:

∆E/W − n · Eloop /W
L= (A.20)
n(L/Dexp,2 + L/Dexp,4 )

166
A.2.8 Traffic Pattern Trajectory Generator

Inputs

• E/W: current aircraft energy

• LS.E td, Etd /W : desired touchdown energy

• r: nominal turn radius, design constant

• L: length of the baseline traffic pattern leg, design constant

• L m: undershoot margin, determines the position of the target touchdown point past
the runway threshold, design constant

• LS.dir: direction of the traffic pattern, either right or left

• TP.L/D exp[]: expected glide ratio for each segment of the traffic pattern in the
order of downwind leg, turn 1, base leg, turn 2, final leg

• r: nominal turn radius

Outputs

• TP.leg[].fix: array of the leg fixes in local North-East coordinate system

• TP.leg[].course: course of each leg in local North-East coordinate system

• TP.leg[].length: length of each traffic pattern leg in m

Trigger

called by higher level state machine, runs once upon exit of holding pattern

167
Description

This function takes the energy error between the current energy level, E/W, and the energy
required to fly the baseline traffic pattern, Etp /W , and elongates or shortens the trajectory
by moving the base leg fix along a line angled at 45◦ to the base leg by magnitude d, which
is dependent on energy error ∆E/W and L/Dav . The energy error is:

∆E/W = E/W − Etp /W − Etd /W (A.21)

where

L−r rπ L − 2r rπ L − r + Lm
Etp /W = + + + + (A.22)
L/Dexp,1 2 · L/Dexp,2 L/Dexp,3 2 · L/Dexp,4 L/Dexp,5

where L/Di are the expected glide ratios on the downwind leg, turn to base leg, base leg,
turn to final leg and final leg from array TP.L/D exp[]. The average glide ratio for the
complete traffic pattern can be found from:

(L − r) · L/Dd + (L − 2r) · L/Db + (L − r + Lm ) · L/Df + 12 rπ · (L/Dt1 + LT 2 )

L/Dav =
3L − 4r + Lm + rπ
(A.23)
The amount, d, by which the base leg fix should be moved is given by:

L/Dav
d = 5.0 · · ∆E/W (A.24)
L/DP OH

This formula was derived from a graph of the d vs ∆E/W for various L/Dav as shown in
Figure A-8. The graph results from comparing the length of the baseline traffic pattern with
the length of the traffic pattern achieved for d ranging from -3,000 ft to + 3,000 ft. For the
detailed equations, refer to module Traffic Pattern Trajectory Updator in section A.2.11.

For L/Dav = 10, the maximum and minimum energy error, with which the holding pattern
can be exited is + 12 Eloop /W = +433f t or − 12 Eloop /W = −433f t respectively. From equation
A.24, the distance to move the base leg fix for energy error +/- 433 ft and L/Dav = 10 is

168
 dT_F
Joining leg fix, FJ

dT_F
Joining leg fix, FJ
crs dL_F
Final leg fix, FL Joining
dT_F
leg fix, FJ Base leg fix, FB
dT_F
Joining leg fix, FJ
Distance d required to fix !E/W on downwind leg for leg length remaining LD=5,000 ft
3000 Joining leg fix, FJ crs dL_F
dT_F Base leg fix, FB
Final leg fix, FL crs dL_F Expected touchdown point, TD
Base leg rfix, FB
L/D = 8 crs dL_F
Base leg fix, FB
crs dL_F
L/D = 9 Final leg fix, FL Base Expected
leg fix, FBtouchdown point, TD r
Final leg fix, FL r
L/D = 10 Final leg fix, FL
2000 Expected touchdown point, TD r dL_B
L/D = 11 Final leg fix, FL r
Expected touchdown point, TD
Expected touchdown point, TD
dL_B
Expected touchdown point, TD dT_B
dL_B

Distance d required to fix error in ft

dL_B
1000
dL_B Position, P
dT_B
dT_B
dT_B
Downwind leg fix,
Position,
FD P
dT_B
Initial point, IP
Position, P
0 Position, P
Initial point, IP on
Downwind leg fix,
Position, P New base leg fix
FB_new FD
Downwind leg fix,
Downwind
th
FD leg fix,
Downwind legIPfix, Figure 24: Definition of intercepts FD and di th
Initial point,
−1000 Figure 24: Definition FD
Initialof intercepts and distances New base leg fix
needed to update trajectory onTp
point, IP
NewFB_new
base leg fix The co s
Initial point, IP once the
The strategy New base
FB_new for shortening the trajectory
leg fix once whilea di la
Initial point, IP
New base
The leg fix for shortening the trajectory
strategy FB_new
while on the base the the p
once a lack of energy has leg been is to begin cutting
detected. This
the the
ensu
prev co
le
onceFB_new Figure 24: Definition
has beenofdetected. intercepts andensures
distances theneeded to isupdate the in
a lack of energy the previous This calculation, finaldL leg F and dLtrajectory
intercepted B are asknown
the
on
earlypo
inter T
b
theFigure Figure 24:
previous Definition of intercepts andknown.distances Byneeded totheupdate trajectory on
pointbas
cou
−2000 24: calculation,
Definition ofdL F and the
intercepts dLand are
B intercept
distancespoint needed shifting
to to final original
theupdate leg IF can
trajectory baseon leg
be fix
deter
base
points F
dle
Figure 24: Definition of intercepts and distances the intercept
The strategy point
neededfor to the
toshortening final
update trajectory leg I
the trajectory can
F on base be determined.
while
leg the on the Together
base of legthe
is tolegwith
begin the current
cutting cours
the dis
posi
cor
points define course joining the
course base ch
The
points
once astrategy
define
lack of fortheshortening
energy coursehas the of
been thedetected.
the trajectory
leg joining Thiswhile
the baseon the
ensures legthe base
and leg is leg.
finalleg
final toisto
begin
This cutting
interceptedcourse the
now distan
corne
defin
leg
The strategy for shortening trajectory course while change
on the from
base legbase toleg
is leg begin joining
cutting legas
the early
and
distance
corner froF
ima
F_
once course lackchange
a previous of energy from has base been legdetected.
to
andjoining This leg ensures
and from the joining
final isto
leg the the
intercepted finalbaseleg.
as whichWith
early leg as
ThFfr
th pB
The strategy for shortening the trajectory once athe
while on the
lack
distance of basecalculation,
energy
dT
leghas
and dT
is been
to begin dL Fcutting
detected.
can be leg
dL the
This
calculated
B are
distance corner
ensuresknown.
which
dT the
B and
immediately By
final
represent
shifting
dT F can
leg is
thethe
be original
intercepted
turn
calculated
anticipation as early leg
leg asfix
from
The
rep
poss tdn
−3000 thethe previous
intercept calculation,
Bpoint to dL
Fthe and
Ffinal dLIBleg arecanfromknown.
be By
determined.
the base shifting
leg Together
and original
with
intercepting basecurrent
the legfor
the fixinterce
F
posit
final B le b
−500 −400 −300 −200 −100 once0 a lack100 of energy 200has been
300 detected. 400 This the500ensures
previous
leg the
from finalbase
calculation,
the leg isleg
dLintercepted
Fand dLBasare
andintercepting early
F known. as possible.
the By leg
final shiftingFromthe
from the original
base leg base leg fixThe
respectively FB(see new
byw”d
thepoints
interceptdefine pointthe to the
course final
ofleg the leg I
legFleg can
joining
The be thedetermined.
base
base leg and Together
final with
leg.hence the
Thiscurrent current
course position
now define
the error
Energy previous
in ft calculation, dLF and dLB are known.
the intercept By
Thedefinenew shifting
pointbase toleg the
the original
fixfinal
FB,new base
Ihence
F can be
becomes: Fnew
fix determined.
B by dLF and
leg fixr,F
Together with
B,new the becomes: position FPB
points
course change the fromcourse of
base the
leg leg
to joining
joining the
leg basefrom
and leg and joiningfinal leg leg.toThis the course
final leg.nowWith defines thi
the intercept point to the final leg IF can points be determined.
define theTogether course ofwith the leg thejoining
currentthe position
base leg P ,and these finaltwoleg. This course now defines the
FB,ne
course change from base leg tobejoining leg and from joining the leg turn
to theanticipation
final leg. WithB,new this k
points define the course of the leg joining the course distance
base leg and
change dT
from B and
final baseleg.dT F can
This
leg to course
joining calculated
nowleg anddefines which
from represent
thejoining
amount legofto the final leg. WithforF intercep
this know .e
distance
leg
F_Bfrom dT.east
Btheand base dTFleg canand be(|intercepting
calculated −which
FB,new the ) represent
.east
final = Ffrom
leg theF
B).east theturn (|anticipation
+base dLleg +r −for
B |respectively dTintercepti· sin(c
B )(see ”g
F
course change from base leg to joining legdistance and fromF dTjoining
B,newB and dT
=FF
leg to .east
can
B thebe +calculated
final leg.B |With
dL +r which dTrepresent
this · sin(crs
Bknowledge, theb theturn .north
anticipation
B,new = F B .north
for intercepting+ (|whdL
legThefrom new thebase baseleglegfixand F intercepting
hence the final leg from the base leg respectively (seewhere
becomes: ”get
distance dTB and dTF can be calculated which leg fromrepresent the turn anticipation
B,new for intercepting
the base leg and intercepting the final leg from the base leg respectively (seewhere the joining ”getTur
Thetdnew base leg fix FB,new hence becomes: cr
Figure A-8: Distance, d, to move base leg fix to correct energy error ∆E/W leg from the base leg and intercepting theThe
The new base leg fix FB,new hence becomes:
final legbase
new fromlegthe
where crsb denotes
fixbaseFB,new leg hence
respectively
the course of the
becomes: (see
where ”getTurnDist.m”).
base crs leg.b denotes the course of theL_D base leg.χb χ
χb
w
Tc
FB,new χb .east
χc = FB .east + (| dLB | +r − dT χbBχ) c· sin(crs
χd b ) F B,new .north = F B .north + The
(| dL B
The fixdetermined
of bthe F J
FB,new .eastThe fix
L_D =F ofBthe.east new dLB | leg
+ (|joining +r − is dTsimilarly
B ) · sin(crs ) Fnew B,new joining
from: .north = legFB is.north
similarly + a(| deter
dL
F|JB jo
.ea |
FB,new .east = FB .east b + (| dLB | +r − dTB ) · sin(crsb ) FB,new .north = FB .north + (| F dLB +r
J .east le
FB,new .east = FB .east + (| dLB | +r − dTB ) · sin(crs F b) F
.east = B,new
F .north
.east + (|=dL FBF.north| +r − +F dT(|F.east
dL) ·B |=+rF−
sin(crs )dTB
.east F )+ · cos(crs
.north(| dL = b )F
| +r −
.north dT + )
(| ·dLsin(crs| +r eq)−
where crsb denotes
J B the course of the base J leg. Bf J F B F F
wh
f
(50)
+/-2,200 ft.
F_D
where tdcrsχbc denotes the course of the base leg.
where crsχbbdenotes ◦ the course of the base leg. whereon
where
cr
χb Theχc χfix d◦45 ofdenotes
the newthe joining legofisthe similarly determined Figure from:8: Energy correction A
where crsb denotes the course of theL_Dbase leg.χb χwhere χ
td
c d 45
crs course final
where determined leg.
crsf denotes the course of the final leg.As so joi
The fix of fthe new joining leg is similarly from: joinin
χb χc χd The fixAs of soon
the new as the aircraft
joining
d leg ishas turned determined
similarly ontosoon
As the joining
as from:
the leg, it will
aircraft has beturned
denotedonto asjoining
base
the leg
joi le
Fjoining
J .east = Fwill
B .east + (|Insufficient
dLF for | +rthe − dT F )leg · sin(crs ) Fdownwind .north = Fbase .north dLF |leg.
+ (| final +r
leg −
ism
Figure A-9 shows the updated trajectory (red curve) compared to the baseline trajectory The fix of the new joining leg is similarly determined
FJ .eastleg = isFBnot
from:
leg
.east
be
at+a (|right
reserved
a FJ .east = FB .east + (| dLF | +r − dTFjoining
| +r −
9. F angle
dL
F_D'
dTF ) · leg
anymore
energy
new
) · sin(crs
with
sin(crs is the
on
connecting
f ) final
not at
fthe
legf )willFJbe
a leg
FJ .north right
Jthe
.north
the
current
reserved = FBfor
equation
=angle
FB .north
Bleg
.north
anymorefor
is
the +leg
+ (|∆L
corrected
new
dL
and
(| dL leg
| +r
FF will
with
connecti
F leg
the
by
−final
need
equationdTto
equ
is−no
| +requat A
dT
F)
FJ .east = FB .east + (| dLF | +r − dTF ) · sin(crsfequation ) FJ .north 52. The = FBprocedure
.north + (| fordL F | +r − dTthe
determining F ) · newcos(crs f)
trajectory is analogous to before.
where crsf denotes the course of theequation final leg. 52. The procedure for determining the ne
whereAs crssoon � (51)
f denotes thethe coursehas of ∆L the finalonto leg. the joining as base leg �
(blue curve) for a runway oriented due East. Figure 8: Energy correction
where crsf denotes the course of the final leg.Asjoining
whereon crsthef denotes
As soon downwind
soon aslegthe
asthe
aswill
aircraft
aircraft
course
∆(E/W
thebeaircraft leg
reserved
of
)=
has
has turned
the
forturned
turned
final
theonto
(L/D)
B leg. ∆LF
new +legthe
onto
b the (L/D)joining
−
joining
connecting
f
Targetleg,
∆L
the
leg,
runway
leg,∆(E/W
2 +
B
it will
it
it∆Lwill
will be
currentbe) =
2 be
F −∆L
base
denoted
2∆L
denoted
(L/D)
denoted leg
B ∆LF · cos(∆crs
B and
asc +
asfinal
base
∆Lleg.
base Fleg As
leg again.
aga
−
joining leg will be reserved for the new leg connecting the current base (L/D)
leg and final (L/D)leg. As f tosob
As soon as the aircraft has turned ontojoining the leg legiswill
joining notleg,
beatitreserved
a will
rightbeangle denoted
for theanymore
new as leg base with legthe
connecting final
again. the leg
The the
termequation
current base leg for and∆L
b F will
final leg. needAs soon
Insufficient joining
energyleg onwill
thebedownwind leg is corrected leg
by ismoving
not at52.
equation atheright
The base angle leg
procedure anymore
inwardsfor with
as
determining the final
shown thein leg
Figure
new the equationisfor
trajectory ∆LF will
analogous to need
before. to be
reserved for the new leg leg connecting
is not atthe current
a right anglebaseanymore
leg and final with leg. the final As soon leg as thethe base for ∆LF will need to be adj
equation
9. equation 52. The procedure for determining the new � trajectory
37 is analogous to before.
leg is not at a right angle anymore with the final leg
equation 52.the Theequation
procedure for for ∆Ldetermining
F will need the to be new adjusted
trajectory as2 inis analogous to before.
equation 52. The procedure for determining the new trajectory is analogous to∆L B
before. ∆LF � ∆LB + ∆LF2 − 2∆LB ∆LF · cos(∆crs)
L_F’ = 7,520 ft L_F= 6,000∆(E/W ft ∆(E/W ) = ∆L B
+ ∆L F � − ∆L 2 + ∆L2 − 2∆LB ∆LF · cos(∆crs)
� )= ∆LB(L/D)+∆L b F
(L/D)− f ∆L2 +B∆L2 −F2∆L(L/D)
B F B ∆LNominal trajectory
F c· cos(∆crs)
∆L2 + ∆(E/W
∆L2F −) 2∆L = (L/D) +b (L/D) −f (L/D)c
∆LB ∆LF B ∆LbF · cos(∆crs)
(L/D) (L/D)f
L_F= 6,000 ft
Target ∆(E/W
runway )= + = 5,610 − ft B (52) (L/D)c Lack of energy
(L/D)b L_F’ (L/D)f !L1 (L/D)c correction
37
37
37
37
Nominal trajectory
L_B’ = 3,610 ft
!E/W = + 433 ft L_B= 5,000 ft !E/W = - 433 ft
Lack of energy
correction
Initial Point

L_B’ = 6,520 ft
L_B= 5,000 ft
Figure 9: Energy error correctio

Initial Point
Initial point, IP Initial point, IP α
L_D’= 3,760 ft !L1
L_D= 5,000 ft Since 45°
moving the base leg parallel to the existing base
19° d=2,200
2 · ∆Lfft. The amount of energy that can corrected for is dire
13° 45° turn onto the base leg would be started. The relationship o
L_D= 5,000
Figure 9: Energy error correction ft
on downwind
to leg
go until the turn is started is given by:
d=2,200 ft
L_D’= 6,700 ft 2·∆
α ∆(E/W ) =
Since moving the base leg parallel to the existing base leg, the length of the trajectory is adjusted by L/
2 · ∆Lf . The amount of energy that can corrected for is directly proportional assuming the same
to the distance to go(L/D) for both legs for simplicity. For
until the
to distance
turn onto the base leg would be started. The relationship of maximum adjustable to goerror
energy untiltothe turn is shown in Figure 10.
distance
to go until the turn is started is given by:
Figure A-9: Baseline and updated trajectory for max. and min. 2 · ∆L
energy error expected f
◦ ∆(E/W ) = (7)
(rotated by −90 from local coordinate system) L/D 12
assuming the same (L/D) for both legs for simplicity. For an (L/D) of 10, this energy correction capability
to distance to go until the turn is shown in Figure 10.

After having deduced the amount, d, that the base leg fix should
12 be moved, the Traffic
Pattern Trajectory Generator fills array TP.leg[] with information on the updated location
of each fix, TP.leg[]fix, the course of each leg, TP.leg[].course, and the length of each leg,
TP.leg[].length, in the local coordinate system. The Traffic Pattern Trajectory Generator
calculates the location of the updated leg fixes based on knowledge of the location of the leg

169
fixes of the baseline traffic pattern, the direction of the traffic pattern and the distance that
the base leg fix is moved, d. The result is:
   
F FD,east 0 −5, 000
 D,north   
   
TP.leg[].fix =  FB,north FB,east  =  −5, 000 − d cos(45◦ ) −5, 000 − d sin(45◦ )  , if LS.dir = ”Left”
   
FF,north FF,east −5, 000 − d cos(45◦ ) 0

   
F FD,east 0 5, 000
 D,north   
   
TP.leg[].fix =  FB,north FB,east  =  −5, 000 − d cos(45◦ ) 5, 000 + d sin(45◦ )  , if LS.dir = ”Right”
   
FF,north FF,east −5, 000 − d cos(45◦ ) 0

The downwind leg fix, FD coincides with the initial point and is not changed during the
traffic pattern trajectory generation process.

The course of each baseline traffic pattern leg is known from the choice of the local North-
East coordinate system. They are:

χd,0 = 180◦ for LS.dir = ”Right” or ”Left”

χf,0 = 0◦ for LS.dir = ”Right” or ”Left”


 90◦ for LS.dir = ”Left”
χb,0 =
 −90◦ for LS.dir = ”Right”

The course of the base leg and final leg are unchanged by the trajectory generation process
and hence coincide with the course of the baseline traffic pattern base leg and final leg. The
new course of the downwind leg can be derived from the location of the new base leg fix
and the location of the downwind leg fix using the tangent function and considering the
quadrant:
�
FB,east − FD,east
χ�d = arctan( �
) (A.25)
FB,north − FD,north

170
Hence,    
χ χ�d
 d   
   
TP.leg[].course =  χb  =  90◦  , if LS.dir = ”Left”
   
χf 0◦
   
χ χ�d
 d   
   
TP.leg[].course =  χb  =  −90◦  , if LS.dir = ”Right”
   
χf 0◦

The length of each leg can be found from geometry:

 � 
(5, 000 + d cos(45◦ ))2 + (d cos(45◦ ))2
 
 
TP.leg[].length =  5, 000 + d cos(45◦ ) 
 
6, 000 + d cos(45◦ )

A.2.9 Energy Curve Manager

Inputs

• TP.leg[].length: length of each traffic pattern leg

• r: nominal turn radius, design constant

• LS.E td, Etd /W : desired energy upon touchdown

• TP.L/D exp[]: expected glide ratio for each segment of the traffic pattern

Outputs

M: array of a struct containing both slope and intercept of each local energy vs. range-to-
go-until-next-segment function in the following format. Refer to Figure A-10 for usage of

171
the symbols.  
bD aD
 
 
 b T 1 aT 1 
 
 
M =  bB aB 
 
 
 b T 2 aT 2 
 
b F aF

 
b au,D
 u,D 
 
 bu,T 1 au,T 1 
 
 
Mu =  bu,B au,B 
 
 
 bu,T 2 au,T 2 
 
bu,F au,F

 
b al,D
 l,D 
 
 bl,T 1 al,T 1 
 
 
Ml =  bl,B al,B 
 
 
 bl,T 2 al,T 2 
 
bl,F al,F

Trigger

started and stopped by higher level state machine, executes every 1 s on receipt of updated
TP.L/D exp and TP.leg[].length

Description

The slope b and intercept a specify slope and intercept of a local energy vs. range to go
curve as shown in Figure A-10. Each local, linear equation can be written as:

fi (R) = bi · Ri + ai = E/W (A.26)

172
where b is the reciprocal of the expected glide ratio, TP.L/D exp[i], on the leg specified
by subscript, i. For example, bF , the slope on the final leg can be found from L/Dexp,f =
TP.L/D exp[1] using the array indices as defined in table 3.2.

1
bF = (A.27)
L/Dexp,f

a denotes the desired energy at the end of the leg specified by the subscript, i, and can
be found by working from left to right on the E/W vs. R curve starting at the desired
touchdown energy, Etd /W as follows:

aF = (Etd /W ) (A.28)

aT 2 = (Etd /W ) + bF · LF (A.29)

and so on. The slopes, bi , and intercepts ai are illustrated in Figure A-10.

Local coordinate system

E/W while on downwind leg

Local coordinate system

while on base leg b_D

a_D
b_T1
a_T1
b_B

a_B
b_T2
a_T2
b_F
a_F Range
Final leg Base leg Downwind leg to go, R
Turn onto final Turn onto base

Figure A-10: Local energy vs. range curve function representation

In addition to the energy vs. range curve itself, the Energy Curve Manager also maintains
a representation of the upper and lower energy boundaries which are shown in Figure A-11.
The tolerance bounds are set in a way that the aircraft may at any point in time have x=5%
more energy than needed to reach the turn onto the final leg and y=3% less energy than

173
 E/W
Expected aircraft
performance

slope = 1/(L/D)_exp E_req/W

Upper energy bound

slope = (1+x/100)/(L/D)_exp

Low energy boundary

slope = (1-y/100)/(L/D)_exp

E_td/W
Range
Final leg Base leg Downwind leg to go
Turn onto final Turn onto base

Figure A-11: Energy error tolerance bounds

required to reach this turn. The tolerance bounds are stored similarly to the target energy
vs. range-to-go curve as an array of slopes and intercepts. The subscript u refers to the
”upper” boundary curve, and the subscript l refers to the ”lower” boundary curve.

fu (R) = bu,i R + au,i (A.30)

fl (R) = bl,i R + al,i (A.31)

A.2.10 Traffic Pattern Energy Error Detector

Inputs

• TP.currSeg: counter indicating the current segment of the traffic pattern trajectory
the aircraft is on

• E/W: aircraft energy normalized by weight

• M: array of slopes and intercepts for each leg as returned by ”Energy Curve Manager”

174
 • M u: array of slopes and intercepts for the upper energy boundary as returned by
”Energy Curve Manager”

• M l: array of slopes and intercepts for the lower energy boundary as returned by
”Energy Curve Manager”

• TP.R: range to go until the start of the next segment

Outputs

• ∆E/W : energy error in m

Trigger

called by higher level function, Energy Management, executes every 1 s

Description

This function uses input TP.currSeg to decide, which array element of M u and M l is
applicable to determining a potential energy error. After having extracted the correct slopes
and intercepts from M, M u, and M l, the Energy Error Detector determines if the current
energy is above or below one of the boundaries. If the current energy is in within the
boundaries, the Energy Error Detector returns ∆E/W = 0, which signals the higher level
Energy Management function that no update is required. If the energy is above the upper
boundary or below the lower boundary, the Energy Error Detector returns the magnitude
and sign of the error, negative representing lack of energy, in output ∆E/W .


 0 if bl,i R + al,i ≤ E/W ≤ bu,i R + au,i


∆E/W = E/W − (bR + a) if E/W > bu,i R + au,i



 E/W − (bR + a) if E/W < b R + a
l,i l,i

175
A.2.11 Traffic Pattern Trajectory Updator

Inputs

• TP.currSeg: current segment the aircraft is on

• ∆E/W : energy error

• TP.L/D exp[]: expected glide ratio for each segment of the traffic pattern

• r: nominal turn radius, design constant

• TP.leg[]: fix, course and length of each traffic pattern leg

• TP.L, L: distance to go until the next leg fix

Outputs

• TP.leg[]: updated fix, course and length of each traffic pattern leg

• TP.updateFlag: sets this flag, if a trajectory update has been performed to signal
the Traffic Pattern Follower to update the active leg within the FMS

Trigger

called by higher level function, Energy Management; executes whenever there is a non-zero
energy error from the Energy Error Detector

Description

This function takes input TP.currSeg and checks whether a trajectory update is currently
allowed. Trajectory updating is possible only on the downwind leg and the base leg with the
current design.

176
Energy error correction on downwind leg

While on the downwind leg, energy errors are corrected by moving the downwind leg fix
along course χc1 which is placed at a 45◦ angle to the course of the base leg. The trajectory
adjustment strategy while on the downwind leg was illustrated in Figure 2-14 in chapter
2.1.9. In order to correct energy error ∆E/W , the base leg fix, F B would have to be moved
along course χc1 by magnitude d. The direction in which F B is moved depends on the sign
of d. A positive d moves the base leg fix outwards, which corrects a positive energy error.
The final leg fix is consequently moved by d · cos(45◦ ) along the opposite direction of the
final leg course.

The sign and magnitude of d can be found from the energy error ∆E/W , the distance
to the base leg fix, L, the angle the downwind leg is making with χc1 , αp and the average
glide ratio to be expected on the remaining trajectory L/Dav . As mentioned previously,
no explicit form for d in terms of ∆E/W , L and αp can be found. However, the implicit
equation can be solved for ∆E/W for vectors of d, L and αp and the result stored in a lookup
table. The implicit equation for d can be found from comparing the energy required to fly
the baseline trajectory to an elongated or shortened trajectory E/Wtp through d.

L� + LB + 2d cos(45◦ ) − 2td + r · ∆χ L + LB − 2td,0 + r · ∆χ0

∆E/W = E/Wtp −E/Wtp,0 = −
L/Dav,tp L/Dav,tp0
(A.32)
with L’ being the distance from the current position to the new base leg fix, td the distance
before the base leg fix that the turn to intercept is started, and ∆χ and ∆χ0 being the
updated and baseline course change from the downwind leg to the base leg. To simplify the
calculation, it is assumed that L/Dav,tp ≈ L/Dav,tp0 = L/Dav . Hence,

1
∆E/W = (L� − 2t�d + r · (135◦ − α) + 2d cos(45◦ ) − (L − 2td + r · (135◦ − αp ))) (A.33)
L/Dav

177
 L/Dav can be written as:

L(L/D)d + LB (L/D)b + LF (L/D)f

L/Dav = (A.34)
L + LB + LF

t�d and td can be found from α and αp respectively:

�
1 + cos γ
td = r (A.35)
1 − cos γ

where γ = 180◦ − ∆χ = 135◦ − α

1 Equations
α can be found from αp :
L sin(αp )
) α = arcsin( χc (A.36)
L� αp (1)
� � αp (2)
where L� = L2 + d2 − 2Ld cos(180◦ − αp ) = L2 + d2 + 2Ld cos(αp ) α

The result can be graphed on d vs. E/W plots, while first holding αp constant and L/

substituting a discrete set of L. This is illustrated in Figure A-12.

d vs.!E/W for L/Dav =10,"old=45° and various L

500

L=1500ft
400 L=3000ft
�
L=4500ft FF,no
300
Distance d that base leg fix has to be moved in ft

�
200 FF,

100

0
χd
−100
αp = 15◦
αp = 45◦
−200 αp = 75◦
L=4500 ft
−300
L=3000 ft
2L=1500Design
ft
Concept
−400

-Engine on guidance to Initial Point -E

−500
−100 −80 −60 −40 −20 0 20 40 60 80 100 -Use Holding pattern -Use trajectory u
Energy error !E/W in ft

3 Functional requireme
Figure A-12: d vs. E/W plot for various L with αp = const = 45◦
The autoland function is designed to al
throttle commands) perform an emerge
178 autoland needs to employ several funct

1. Select destination
2. Guide the aircraft to an initial po
3. Hold the aircraft on a holding p
within a band of energy required
 χc
α α (1) L/Dav =
p
αp (2)
α L/D

� L
FB,nor
For the second plot, L is held constant and αp takes on a set of discrete values. This is
�F
FB,ea
illustrated in Figure A-13. All of this information is stored in a lookup table for L/Dav = 10.
�
FF,north =F
d vs.!E/W for L/Dav =10,L=5000ft and various "old
500
�
"old=15° � FF,nor
400
°
FF,east =
"old=45

"old=75°
300 ��
Distance d that base leg fix has to be moved in ft

FFF,e
F,
χ
200
χd
FF
100
αp = 15◦
◦
0
αp =χ45
d
◦
αp =α75
p = 15
◦
−100
αp χ= 45◦
d ◦
αp α= 75
= 15◦
−200
2 Design
p Concept
◦
−300
αp = 45
-Engine
5000 fton
L= 2 = 75◦ to
αpguidance
Design Initial Point -Engin
Concept
−400 in each case
-Use Holding pattern -Use trajectory updati
-Engine on guidance to Initial Point -E
−500
−150 −100 −50 0 50 100 150 2 Design Concept
-Use Holding82pattern -Use trajectory up
Energy error !E/W in ft
3 Functional requirements
-Engine on guidance to Initial Point
The 3 -Use
autoland Holding
function
Functional pattern -Use trajectory
is designed to allow a
requireme
Figure A-13: d vs. E/W plot for various αp with L=constthrottle
= 5,000 ft
commands) perform an emergency la
autoland needs to function
The autoland employ several functions.
is designed to allo
3 Functional
throttle commands) perform
requirem
an emergen
To find d for a specific combination of ∆E/W , L and αp , the data in 1.the lookup
Select table is
destination
autoland needs to
The autoland employisseveral
function designedfuncti
to a
interpolated using a linear interpolation function that is called by passing
2. the known
Guide
throttle vectors
thecommands)
aircraft to perform
an initialanpoint
emergo
1. Select destination
autoland needs to employ several fun
of ∆E/W , L and αp , the lookup table, which is stored as a 3D matrix and theGuide
3. Hold
2. combination
the aircraft on a holding pattern
the aircraft to an initial poi
within a banddestination
1. Select of energy required to co
of ∆E/W , L and αp for which d should be found. The value d0 returned from the table
3. Hold the aircraft is on a holding pa
4. Guide
2. the
Guideaircraft
the along the
aircraft traffic
to an pa
initial
within a band of energy required t
energy deviates
scaled for an L/Dav of 10. Hence, the actual d for the actual L/Dav can be found from: from the expected ene
4. 3. Holdthe
Guide theaircraft
aircraftalong
on a the
holding
traffi
5. Perform a flarea maneuver
within bandfrom as therequire
of energy aircra
energy deviates the expected
L/Dav
d = d0 5. 4. Guide a(A.37)
Perform the aircraft
flare alongasthe
maneuver thetra
a
10 energy deviates from the expect
5. Perform a flare maneuver as the
When calculating d to move the base leg fix, it is necessary to observe whether the fix
can be moved by the amount required without violating the condition that the course of
the base leg remains unchanged. For a positive d, i.e. elongating the trajectory, this is
not a problem. When shortening the trajectory, however, the magnitude of d is limited by
L(sin(135◦ −αp )
dmax = sin(45◦ )
from geometry. Any remaining energy error will therefore be carried
over to the base leg, where it can be corrected for using the following strategy. Refer to
function ”Traffic Pattern Generator” for details on how to update the traffic pattern leg

179
fixes, TP.leg[].fix, the courses, TP.leg[].course and the length of each leg, TP.leg[].length.

180
 Energy error correction on base leg

While on the base leg, energy errors are corrected by inserting a fourth leg, called the ”Final
Leg Intercpet Leg” as shown in Figure 2-15 in chapter 2.1.9 and moving the final leg intercept
1 Equations
fix, FF I along course χc2 , which is placed at a 45◦ angle to the course of the base leg. The
d0 = f (∆E/W, L, α
final leg fix, FF is moved to a fixed position 500 ft off the runway threshold. During future
updates, only the final leg intercept fix is moved while the location of the final leg fix remainsd = d L/Dav
0
10
unchanged as illustrated in Figure A-14, which χb repeats Figure 2-16 from chapter 2.1.9 for
convenience. Due to the added leg, the distance αp (1)
that the final leg intercept fix should be
αp (2)
moved, d, is not only dependent on L and αp as α was the case for the downwind leg, but also
dependent on the distance that the final leg intercept fix has already been moved from the LD (L/D)d + LB
L/Dav =
location of the baseline final leg intercept fix FF I,0 , termed dtotal . Similarly to the downwind LD + LB

αp = χc1 − χd

1 Equations
�
FB,north = d cos(χc1 ) + F
where γ = π− | (χ1 − χ2 ) |.
1 Equations
�
−−−−−−→ χc FB,east
F_FI’ = d sin(χ c1 ) + F
d = f (∆E/W,
dtotal =| FF I,0 , FF I | (A.58)
L_FI χc
1 d Equations
αp
α �
d_total
αp (1) FF,north = d cos(45◦ ) cos(χdc1
0
αp (2)
F_FI’
α L
F_F=F_F’
2 Design
�
FF,east =Concept
d cos(45◦ ) sin(χ
L (L/D
c1
D
L/Dav =
F_FI,0 χb
L_F=1,500 ft -Engine on guidance to Initial Point -
αp (1)
Previous energy corrections
α-Use
p (2) χ�d =-Use
Holding pattern trajectory
atan2(∆E, αp =
∆Nu
χ
Aircraft
created d_total >0 α
position, P
χc2
Next update for lack of energy: 2 Design Concept
-> d negative χc1 3 Functional requirem L/Dav =
◦ -Engine on guidance to Initial Point -Engine off appr
αp = 15
Thepattern
-Use Holding autoland
-Usefunction
trajectoryisupdating
designedfortoene
a
αp = 45◦
throttle commands) perform an emerg
αp = 75◦
autoland needs
3 Functional to employ several func
requirements
Figure A-14: Geometry used to determine difference in path length between the current and �
FB,north
updated trajectory The autoland1.function
Select isdestination
designed to allow an aircraft
2 Design Concept
throttle commands) perform an emergency landing fro
F �
autoland needs to employ
2. Guide theseveral functions.
aircraft to an B,east
initial p
-Engine on guidance to1.Initial Point -Engine off approach
Select destination �
fro
3. Hold updating
-Use Holding pattern -Use trajectory the aircraftforonenergy
aFF,north
holding= dp
man
2. Guide the within
aircraft atoband
an initial point onrequired
of energy the stan
leg, the energy error that can be corrected on the base leg is determined by the difference in
3. Hold the aircraft on a holding pattern
�
FF,east = in
by the d
4. Guide the aircraft along the traffi
3 Functional requirements
within a band of energy required to commence t
energy deviates from the expecte
181 4. Guide the aircraft along the traffic pattern whilχ�d
The autoland function is designed
energy5. to allow
χd deviates
Perform anexpected
fromathe aircraft
flare to as
energy
maneuver semi-
the
throttle commands) perform αan
5. Perform
p =emergency
15 ◦
landing from any p
a flare maneuver as the aircraft reaches
α 6.
= Capability
45 ◦ to absorb errors in w
autoland needs to employ several
p functions.
6. Capability to◦ absorb errors in wind measuremen
αp = 75
7. Capability to absorb errors in L/
1. Select destination 7. Capability to absorb errors in L/D0 estimation
the length of trajectory:

1
∆E/W = (L� +L�F I −2t�f i +r ·(180◦ −(α+β)))−(L+LF I −2tf i +r ·(180◦ −(αp +βp )))
L/Dav
(A.38)

L/Dav can be written as:

L(L/D)b + LF I (L/D)f i
L/Dav = (A.39)
L + LF I

t�f i and tf i can be found from α and αp respectively:

�
1 + cos γ
tf i = r (A.40)
1 − cos γ

where γ = 180◦ − ∆χ = 180◦ − (α + β)

α can be found from αp :

L sin(αp )
α = arcsin( ) (A.41)
L�
�
where L� = L2 + d2 − 2Ld cos(αp )

β can be found from βp :

LF I sin(βp )
α = arcsin( ) (A.42)
L�F I
� �
where L�F I = L2F I + d2 − 2LF I d cos(βp ) with LF I = L2I + d2total − 2LI dtotal cos(135◦ ) and
◦
βp = arcsin( LI sin(135
LF I
)
).

The result can be graphed on d vs. E/W plots, while first holding dtotal and αp constant
and substituting a discrete set of L. This is illustrated in Figure A-15.

For the second plot, L and dtotal are held constant and αp takes on a set of discrete values.
This is illustrated in Figure A-16.

Finally, L and αp are held constant while dtotal takes on a set of discrete values. Figure
A-17 illustrates this.

182
 d vs.!E/W for L/Dav =10,"p=30°, dtotal=500 and various L
500

L=2000ft
400 L=2500ft
L=3000ft
300
Distance d that final leg fix has to be moved in ft

1 Equations
200 !p=30°
d = f (∆E
100 1 χ Equations
c
αp (1) L=2000 ft
0 d = f(
αp (2)
−100 χc α

−200
αp (1)L=2500 ft LD
αp (2) L/Dav =
−300
1 α L=3000
Equations
ft
αp
−400
d =avf (∆
L/D =
−500 �
−50 −40 −30 −20 −10 0 10 20 30 40 50 χc FB,north =
Energy error !E/W in ft
αp (1)
αp (2) �
FB,east =
α
Figure A-15: d vs. E/W plot for various L with dtotal =500 ft and αp = const = 30◦ � F�
FF,north B,north
= d cos
L
L/Dav =
�
�
FF,east F
=B,eas
d cos
α
d vs.!E/W for L/Dav =10,L=3000ft, dtotal=500 and various "p
500
� χ�d =
FF,north = ad
"p=15° �
χd FB,north =
400 "p=45° αp = 15◦ �
"p=75° αp = 45◦
FF,east =d
�
300 FB,east =
Distance d that final leg fix has to be moved in ft

αp = 75◦ L

200
�
χ�d
FF,north = dc
100
2χd Design Concept
◦
αp = 15
-Engine on guidance to Initial PointF-Engine
� off
0
αp = F,east = d c
45◦ pattern -Use trajectory updating
-Use Holding fo
−100
αp = 75◦ L
χ�d =
−200
3 Functional requirements
χdDesign Concept
2 The
−300
= 15◦ function is designed to allow an airc
α autoland
p
throttle
αp = oncommands)
45◦ perform an emergency landin
-Engine
autoland
guidance
needs
◦
to Initial Point -Engine
to employ several functions.
−400 α p = 75
-Use Holding pattern -Use trajectory updatin
L
−500
1. Select destination
−60 −40 −20 0 20 40 60
Energy error !E/W in ft
32 2.Design Concept
Functional requirements
Guide the aircraft to an initial point on the
3. Hold
-Engine the aircraft
on guidance to on a holding
Initial Pointpattern
-Enginebyot
The
-Useautoland function
within apattern
Holding is designed
band of-Use
energy toupdating
required
trajectory allow
to an
commen
Figure A-16: d vs. E/W plot for various αp with dtotal = 500f t and L=
throttle 3, 000f tperform an emergency lan
commands)
4. Guide
autoland theto
needs aircraft
employ along the traffic
several pattern
functions.
energy deviates from the expected energy
3 Functional requirements
1.5.Select destination
Perform a flare maneuver as the aircraft rea
For the trajectory updating strategy on the base leg, it is possible that the course
The autoland change
function is designed to allow an ai
2. Guide
throttle the aircraft
commands) to an
perform an initial pointland
emergency on
◦
from the updated base leg to the final intercept leg is greater than 135 . This
autoland
3. Hold the will
needs to cause
employ
aircraft
several functions.
on a holding pattern
the FMS to cap turn anticipation in favor of overshooting the subsequent leg. This avoids 1. within a band of energy required to com
Select destination
4.
2. Guide the
the aircraft
aircrafttoalong the traffic
an initial patt
point on th
energy deviates from the expected energ
183 3. Hold the aircraft on a holding pattern by
within a band
5. Perform a flareofmaneuver
energy required
as thetoaircraft
comm
4. Guide the aircraft along the traffic patter
energy deviates from the expected energy
5. Perform a flare maneuver as the aircraft r
 χc
αp (1) �
αp (2) FF,north =
α
� �
FF,east
FB,n=
L
FχB�
χd
�
500
d vs.!E/W for L/Dav =10,L=3000ft, "p=45° and various dtotal αp = 15◦ FF,north
dtotal=−1500 ft
d_total=-500ft αp = 45◦
400 dtotal=0 ft αp = 75◦ �
dtotal=1500 ft
FF,east
300 L
Distance d that final leg fix has to be moved in ft

200 2 Design Concept

�
FF,
χd on guidance to Initial Point -Engin
100 -Engine
αp = 15◦pattern -Use trajectory updati
-Use Holding
0
αp = 45L◦ FF
−100 d_total=0 ft αp = 75◦
3 Functional requirements
−200

−300
2 autoland
The Design
χd Concept
function is designed to allow an
throttle commands) ◦ perform an emergency la
αp = 15
-Engineneeds
autoland on guidance
to employto several
Initial Point -Eng
functions.
−400
αp = 45◦
-Use Holding pattern
◦ -Use trajectory upda
−500
αpdestination
1. Select = 75
−50 −40 −30 −20 −10 0 10 20 30 40 50
Energy error !E/W in ft L

d_total=500ft
32. Guide
2Functional
the aircraft torequiremen
Design Concept
an initial point o
3. Hold
The the aircraft
autoland functiononis adesigned
holding to
pattern
allow
-Engine
within a on guidance
band of energy torequired
Initial Point
to co
throttle commands) perform an emergency
◦ -Use Holding pattern -Use trajectory
Figure A-17: d vs. E/W plot for various dtotal with αp = 45 and 4.
L= 3, 000f
autoland
Guide needs tto employ
the aircraft several function
along the traffic pa
energy deviates from the expected ene
1. Select destination
3 Functional requirem
5. Perform a flare maneuver as the aircra
2. Guide the aircraft to an initial point
The autoland function is designed to a
3.throttle
Hold thecommands)
aircraft onperform an emerg
a holding patte
autoland
within a needs
band to employ required
of energy several func
to
excessive corner cutting. Figure A-18 shows the effect of this scheme on the required d for
4. Guide the aircraft
1. Select along the traffic p
destination
a given ∆E/W . While d increases rapidly with increasing ∆E/W before turn anticipation
energy deviates from the expected en
2. Guide the aircraft to an initial p
is capped, it increases noticeably slower after the capping, which makes intuitive sense
5. Perform from
a flare maneuver as the airc
3. Hold the aircraft on a holding
the illustration shown on the right of Figure A-18. On the other end, a large negative
within ad band
will of energy require
at some point cause an elongation of the trajectory rather than a shortening,4.which
Guide occurs
the aircraft along the tra
energy deviates from the expect
as soon as the course of the updated base leg becomes equal to the course of5.the updated
Perform a flare maneuver as the
final leg intercept leg. This point signifies the maximum lack of energy, ∆E/W , that can
be corrected using the trajectory updating strategy on the base leg. As with the downwind
leg, all this data is stored in a multi-dimensional array, which acts as a lookup table, and is
used to calculate an interpolated value of d for any combination of ∆E/W , L, αp and dtotal .
The value returned, d0 is again scaled for an L/Dav of 10, so that d can be found from:

L/Dav
d = d0 (A.43)
10
184
 L/Dav =

αp = χ

�
FB,north = d co

max. corner cutting reached (!">135°) �

FB,east = d sin
1
aircraft overshoots next leg EquationsF_FI ’ +

d vs.!E/W for L/Dav =10,"p=30°, dtotal=500 and various L

�
2500
FF,north cos(45◦
d ==fd(∆E/W,
L=2000ft
2000 L=2500ft
L=3000ft
F_F F_FI �
1500 FF,east = d cos(45
d=
Distance d that final leg fix has to be moved in ft

1000 χb
αp = 30◦ χ�d = atan
500 F_FI-’
αp (2)
χc2 L=2000 ft
0 α
L=2500 ft
χc1
−500
αp = 15◦
max. –ve energy
L=3000 ft LB (L/D
correction reached
L/Dav =
−1000 αp = 45◦
increasing d, elongates
L
αp =of 75◦
trajectory instead
−1500
shortening it αp =
−2000

−2500
2 Design Concept �
FB,north = dc
−200 −150 −100 −50 0 50 100 150 200 250 300
Energy error !E/W in ft
-Engine on guidance to Initial Point -Engine
�
off app
FB,east
-Use Holding pattern -Use trajectory updating = en
for ds

F,north F� = d cos(4
Figure A-18: d vs. E/W plot for extreme ∆E/W to3 illustrate corner cases
Functional requirements
�
FF,east
The autoland function is designed to allow an=aircraft
d cos(4
throttle commands) perform an emergency landing fro
From knowledge of d and the location of the current leg fixes, the leg
autoland fixes
needs can beseveral
to employ updated
functions. χ�d = ata
as follows: χ
1. Select c2destination
χc1
   2. Guideαpthe
= 15◦
aircraft  initial point on the sta
to an
◦
F FD,east 0 α−5,
p = 45
000
 D,north   3. Hold αthe 75◦
p = aircraft on aholding pattern by the
 � �    required to commence
 FB,north FB,east   Pnorth within aPband 
east of energy

TP.leg[].fix =  =  , if LS.dir = ”Left”
dtotal =
  
 FF� In orth FF� I,east ◦ ◦
  FF I,north − d cos(45 ) FF I,east + d sin(45 ) 
   
2 Design Concept
FF,north FF,east 0 −500
-Engine on guidance to Initial Point -Engine off ap
-Use Holding pattern -Use trajectory updating for

   3 Functional requirements
F FD,east 0 5, 000
 D,north   The autoland function is  designed to allow an aircra
 � �   throttle  an emergency landing
commands) perform
 FB,north FB,east   Pnorth Peast 

TP.leg[].fix =  = autoland needs to employ  , several functions.
   if LS.dir = ”Right
 FF� In orth FF� I,east ◦
  FF I,north − d cos(45 ) FF I,east ◦
− d sin(45 ) 
   1. Select destination
FF,north FF,east 0 −500the aircraft to an initial point on the s
2. Guide

The base leg fix, FB will be changed to the current position. The final intercept leg fix,
FF I is moved along course χc2 as shown in Figure A-14 and the final leg fix, FF remains

185
unchanged.
Note: If this is the first update from the baseline traffic pattern trajectory, the final leg
intercept leg needs to be inserted first and FF moved to the fixed position 500 ft off the
runway threshold.

The updated course of the base leg and final intercept are derived from the location of
the new base leg fix, FB� , the final leg intercept fix, FF� I and the final leg fix, FF . The final
leg course is unchanged by the trajectory generation process. The downwind leg course is
irrelevant at this point and left as is for simplicity. In summary,
   
χ χd
 d   
   F� −F � 
 χb   arctan( F � F I,east −FB,east
� ) 
TP.leg[].course = 

=
 
F I,north
F −F �
B,north 

 χf i   arctan( F � F,east −FF� I,east ) 
   F,north F I,north 
χf 0 ◦

The length of each leg can be found from geometry:

 
LD
 � 
 
 (Ff i,east − Peast )2 + (Ff i,north − Pnorth )2 

TP.leg[].length =  � 

 (Ff,east − Ff i,east )2 + (Ff,north − Ff i,north )2 
 
1, 500

186
A.3 Modules of Layer Trajectory Following

A.3.1 Trajectory to Initial Point Follower

Inputs

• TI.leg[].fix, Pi : array of waypoints specifying the trajectory to the initial point in the
local North-East coordinate system

• LS.hdg, χ − r: runway heading, i.e. approach direction, with respect to true north

• LS.loc, µ0 ,ι0 : latitude and longitude of the runway threshold or location of the earliest
desired touchdown point for alternate landing site

• state.loc: latitude and longitude of current aircraft position

Outputs

• TF legs/ DF leg: track-to-fix leg for trajectory to the initial point, direct-to-fix leg
for direct flight to an unpopulated area

• TI.L: distance to go until the next leg fix

• TI.currSeg: counter indicating the current leg of the trajectory to the initial point
the aircraft is on

Trigger

started and stopped by higher level state machine, executes continuously while active

187
Description

The function of the Trajectory to Initial Point Follower is to translate the trajectory given
in local North-East coordinates to the global latitude-longitude coordinate system using
the runway heading, χr , and location of the runway threshold, (µ0 ,ι0 ). Library function
”convertToGlobalCoords” is used to perform this task. Furthermore, this function keeps
track of the aircraft’s progress along the trajectory. It passes TI.currSeg, indicating the
array index of the current leg, and TI.L, the distance to go until the start of the next leg, to
the Feasibility Monitor. Counter TI.currSeg is incremented every time the aircraft initiates
a turn to intercept the next leg. This initiation can be detected by checking whether the
aircraft has passed the line drawn perpendicular to the current course at the point where
the FMS initiates the turn to allow a smooth intercept to the next leg. This perpendicular
line is found using library function ”getTurnEntryLine”, whereas passing the line is checked
by library function ”lineCrossed”. TI.L is calculated using library function ”calcDistToGo”,
which averages cross-track and along-track error with the straight line distance to the next
waypoint to estimate the length of the actual trajectory to the next waypoint.

A.3.2 Holding Pattern Follower

Inputs

• LS.IP: location of the initial point in local North-East coordinates

• LS.loc: latitude and longitude of the runway threshold or location of the earliest
desired touchdown point for alternate landing site

• LS.hdg: runway heading, i.e. approach direction, with respect to true north

• LS.dir: direction of the traffic pattern, either right or left

• state.loc: latitude and longitude of current aircraft position

188
 • HP.length: length of the straight line segments of the holding pattern

• HP.#loops: number of loops to remain in the holding pattern after completion of the
current loop

• r: nominal turn radius, design constant

Outputs

• HP.R: range to go until the end of the straight line segment, outputs zero during turn
segments

• HP.currSeg: counter indicating the current segment of the holding pattern the air-
craft is on (1=turn to outbound leg, 2=outbound leg, 3=turn onto inbound leg, 4=in-
bound leg)

• HM leg: specification of leg ”hold-until-manual-termination” consisting of initial

point, HM.fix in latitude and longitude coordinates, holding pattern direction, HM.dir,
course of the outbound leg, HM.hdg, and length of the straight line segment, HM.length

• exit: flag indicating to the FMS to exit the holding pattern and to continue onto the
downwind leg, which has been pre-stored

Trigger

started and stopped by higher level state machine, executes continuously when active

Description

The function of the Holding Pattern Follower is to translate the inputs describing the re-
quired holding pattern into the HM leg format required by the FMS and to initiate the
”manual” termination once the Holding Pattern Generator/Updator signals exit from the

189
holding pattern by setting HP.#loops=0. The holding pattern fix, HM.fix is equal to the
initial point, LS.IP. The FMS requires this fix to be specified in global latitude and longi-
tude coordinates, which is performed using library function ”convertToGlobalCoords” . The
course of the outbound leg, HM.hdg is simply equal to the course of the runway heading, so
that HM.hdg=LS.hdg. The holding pattern direction is opposite the traffic pattern direc-
tion, so that HM.dir = -LS.dir. The length of the straight line segment is provided as an
input so that HM.length = HP.length.
The distance to go until the start of the next turn segment can be found from the current
location, state.loc and the knowledge of the two end-points of the straight line segments,
IP and O as shown in Figure A-19. Point IP is known from LS.IP. Point O can be found

Location and direction

OP I of holding pattern for
LEFT traffic pattern

North
L

IP
5,000 ft
O East
2r

Figure A-19: Geometry used to determine location of IP and OP

from knowledge of the traffic pattern direction, the length of the straight line segment L =
HP.length and the turn radius, r.
For a left traffic pattern, the coordinates of O in the local North-East coordinate frame
become:

OPl,north = L (A.44)

OPl,east = −(5, 000f t + 2r) (A.45)

190
For a right traffic pattern, the coordinates are:

OPr,north = L (A.46)

OPr,east = 5, 000f t + 2r (A.47)

The distance to go to IP or OP while the aircraft is on the inbound or outbound leg re-
spectively, HP.R, can be calculated using library function ”calcDistToGo” after the aircraft
position has been converted to local coordinates using library function ”convertToLocalCo-
ords”. The Holding Pattern Follower keeps track of the segment, the aircraft is currently
on by monitoring whether the aircraft has passed point IP, OP, I, and O using library func-
tions ”getTurnEntryLine” and ”lineCrossed”, which first draw a line perpendicular to the
current course through the point of interest and subsequently checks the aircraft’s position
with respect to this line, to determine whether the aircraft has crossed the line. Finally, the
Holding Pattern Follower sets flag ”exit”, which emulates the operator to manually termi-
nate the holding pattern, once input HP.#loops equals 0, and the aircraft has intercepted
the inbound leg. The logic can be written as follows:

 0 if HP.#loops �= 0
exit =
 1 if HP.#loops = 0& HP.currSeg = 4

A.3.3 Traffic Pattern Follower

Inputs

• LS.loc: latitude and longitude of the runway threshold or location of the earliest
desired touchdown point for alternate landing site

• LS.hdg: runway heading, i.e. approach direction, with respect to true north

• TP.leg[]: array of trajectory legs containing waypoints,TP.leg[].fix and course of each

leg, TP.leg[].course

191
 • TP.updateFlag: flag indicating whether the Traffic Pattern Trajectory Updator has
changed the current trajectory

• state.loc: latitude and longitude of current aircraft position

• r: nominal turn radius, design constant

Outputs

• FM leg: specification of single ”fix-to-manual termination” legs which require infor-

mation of the course to track, FM.hdg, and the start fix, FM.fix

• TP.R: range to go until the start of the next segment

• TP.currSeg: counter indicating the current segment the aircraft is on (1=downwind

leg, 2=turn to base, 3=base leg, 4=turn to final intercept leg (if applicable), 5=final
intercept leg (if applicable), 6=turn onto final, 7=final leg)

• TP.L: distance until the next leg fix

Trigger

started and stopped by higher level state machine, runs continuously while active

Description

This function passes one FM leg at a time to the FMS for tracking and times the switching
from one leg to the next, so that a smooth intercept of the next leg is possible without
overshoot. The distance from the next fix, at which the turn should be initiated, is de-
termined by the course change from the current to the next leg, ∆χ, which can be found
from TP.leg[].course and the nominal turn radius, r. This value is found using library func-
tion ”calcTurnDist” and is termed td . Using, td , library functions ”getTurnEntryLine” and

192
”lineCrossed” are used to determine whether the aircraft has passed the point, at which the
turn should be initiated. Furthermore, the Traffic Pattern Follower passes updates of the
current leg, should the Traffic Pattern Trajectory Updator raise the TP.updateFlag.
Note: The energy error tolerance bounds were chosen in such a way that updates do not
occur more frequently than every 3-5 s if the performance estimation error remains within
+/-10%, which is deemed sufficient time for the FMS to capture an updated leg. If, however,
updates occur more frequently, this function will need to include a timer that is started after
each update and expires after 3-5 s. During this time, updates are discarded.

In addition to that, the Traffic Pattern Follower passes TP.R, the distance to go until
the start of the next segment, and TP.L, the distance to go until the next waypoint to the
Energy Curve Manager and Traffic Pattern Trajectory Updator respectively. TP.R is found
using library function ”calcDistToGo” by passing the current location, state.loc, converted
to local coordinates and point Z, which is located at distance td from the next fix, F , along
the opposite of the current course, χ as shown in Figure A-22. The coordinates of Z can be
found as follows:

Znorth = Fnorth − td cos(χ) (A.48)

Zeast = Feast − td sin(χ) (A.49)

TP.L is also found using function ”calcDistToGo” by passing the current location, state.loc,
and the location of the next fix, F , both in local coordinates. Finally, the Traffic Pattern
Follower specifies the current leg the aircraft is on through counter TP.currSeg. If no final
intercept leg exists yet, since no updates on the base leg have been performed, there are only
five traffic pattern trajectory segments, downwind leg, turn to base, base leg, turn to final
and the final leg. The first update on the base leg inserts the final leg intercept leg that adds
two new segments: turn to final intercept leg and the final intercept leg itself. The Traffic
Pattern Follower increments TPcurrSeg every time the aircraft passes the line perpendicular

193
to the point at which the turn to the next leg is initiated, i.e. at the point where the Traffic
Pattern Follower passes the new leg, and at the point where the aircraft has completed the
turn and intercepted the new leg. This can be detected by checking whether the aircraft
has passed the line drawn perpendicular to the current course at a distance td from the
current fix along the current course. This perpendicular line is found using library function
”getTurnEntryLine”, whereas passing the line is checked by library function ”lineCrossed”.
TP.currSeg is used by the Energy Error Detector to index the array of energy vs. range to
go curve and boundaries in M, M u and M l. These arrays are automatically updated to
include the final intercept leg by the Energy Curve Manager once this function receives an
updated array of TP.leg[] from the Traffic Pattern Follower.

A.3.4 Flare Controller

Inputs

• h: altitude above touchdown zone

Outputs

• vsink : target vertical speed to be passed to autopilot vertical speed hold controller

Trigger

started and stopped by higher level state machine, runs continuously while active

Description

Autoland performs a flare maneuver similar to the one implemented by autoland systems for
commercial aircraft. It commands a smooth decrease in vertical speed from the start of the
flare until touchdown, with the difference that the final target sink speed does not approach

194
zero, but a constant vtd . Autoland uses the following equation for the target sink speed, v
in terms of time elapsed, t:
t
v(t) = (v0 − vtd )e− τ + vtd (A.50)

where v0 represents the vertical speed at the start of the flare, vtd = 3.3f t/s the target
touchdown speed at the expected touchdown point, and τ = 1s the time constant at which
the vertical speed command is decreased. With this choice, the flare height becomes 70 ft,
which is the height above ground at which the higher level state machine switches to the
”flare” state and invokes this function.

A.4 Selected Library Functions

A.4.1 convertToGlobalCoords

Inputs

• origin, O, µ0 ,ι0 : latitude and longitude of the origin of the local North-East coordinate
frame

• heading, χ: orientation of the local North-East coordinate frame with respect to true
north of the global coordinate system

• pos, P: North and East coordinates of the point P in the local coordinate frame

Outputs

• loc, µ, ι: latitude and longitude of point P in global coordinate frame

Trigger

Function call

195
Description

This function first rotates the local coordinate system to the orientation described by course
χ to convert P to P’:
    
�
Pnorth cos(χ) − sin(χ) Pnorth
 =  
�
Peast sin(χr ) cos(χ) Peast

Subsequently, this function calculates the change in latitude, ∆µ and longitude, ∆ι between
point P’ and the origin of the local North-East coordinate system, at specific latitude µ0
using ∆N and∆E, which represent the difference in North and East coordinates of P’ and
the origin.
∆µ = arctan( R1N )∆N
1
∆ι = arctan( RM cos(µ 0)
)∆E

where RN specifies the curvature of the earth in the prime vertical and RM the curvature
of the earth in the prime meridian at latitude µ0 , which can be found from approximation
formulas or tabulated data. As an approximation, RN and RM may be set to the average
radius of the earth R = 6, 378, 137m. For a distance of 30 miles, as is assumed to be a typical
distance to a runway, the error from neglecting the flattening of the earth would amount to
300 ft at 45◦ latitude, which is considered acceptable considering the included range margin
of 30%. Finally, the location of P in the global coordinate system is found by offsetting ∆µ
and ∆ι by µ0 and ι0 .
µ = µ0 + ∆µ
ι= ι0 + ∆ι

A.4.2 getTurnEntryLine

Inputs

turnDist, t: distance to waypoint at which turn should be initiated to avoid overshoot

196
 P: location of waypoint in local North-East coordinates

χ: course of the current leg

Outputs

I: intercept point at distance t from waypoint P on the current leg

straight line segment in the direction that yields a right-handed coordinate system if the
Q: point at (arbitrarily chosen) distancealong
10 the
from intercept
course point
of the current I on
straight lineline perpendicular
segment

to the current leg in the direction that Description

6.10.3 yields a right-handed coordinate system if the
straight line segment in the direction that yields a right-handed coordinate system
along the course of the current straight line segment
x-axis points along the course ofFinds
the two
course at a
points describing the turn entry line. The turn entry line is the line perpendicular t
current straight
distance
line the
turnDist fromin
segment
targeted waypoint as shown in Figure 16. p1, the interc
corrected on the base leg is determined by the difference
6.10.3 Description
the length of trajectory:
p2 are found as follows.
Finds two points describing the turn entry line. The turn entry line is the line perpendi
1 course at◦ a distance turnDist from
p1.east the targeted
= wpt.east waypoint
− turnDist ◦
◦as shown
· cos(90 in Figure 16.
− crs) p1, the
∆E/W = (L� +L�F I −2t�f i +r·(180 −(α+β)))−(L+L
p2 are found as follows. F I −2t f i +r·(180 −(α
= wpt.east − turnDist · sin(crs) p +β p )))
Description
corrected L/Dav,tp0
on the base leg is determined by the difference in the length of trajectory:
p1.north = wpt.north − turnDist · sin(90◦ − crs)
(4.44)
p1.east = wpt.east − turnDist · cos(90◦ − crs)
= wpt.north − turnDist · cos(crs)
1av,tp0 can� be� written = wpt.east − turnDist · sin(crs)
∆E/W = L/D
Finds two points describing the �fturn
(L +LF I −2t as:entry◦ −(α+β)))−(L+L
i +r·(180 line. The turn entry line
F I −2t fis the line◦ −(α
i +r·(180
p1.north = wpt.north perpendicular
p +βp )))
− turnDist · sin(90◦ − crs)
L/Dav,tp0 Since the slope of the line perpendicular to the current course is defined by m2 = − m1 an
= wpt.north − turnDist · cos(crs) 1
to the current course at a distance t fromL(L/D the targeted waypoint as shown in Figure 1 A-20.
(4.44)
b ) + (LF I + LF )(L/D) f ◦ − crs) =
m1 = tan(90
L/Dav,tp0 = tan(crs) (4.45)
I, the
L/Dintercept
av,tp0 can point and Qas:are found as follows.
be written Since theL slope
+ LFofIthe
+ line
LF perpendicular
m2 = −tan(crs)to the current course is defined by m2 = −
α = atan(−tan(crs)) = −(crs) 1
m1 = tan(90◦ − crs) =
tan(crs)
t�f i and tf i can be found from αInand
L(L/D b ) +αto
order (Lrespectively:
F I +that
p ensure LF p2
)(L/D)
is always
f in the direction from p1 that forms a right-handed coord
L/Dav,tp0 = need Waypoint, P between abs(crs) <= π m
to distinguish 2 =abs(crs)
and > π . (4.45)
−tan(crs)
L + LF<=
For abs(crs) I +
π LF
2 2
α = atan(−tan(crs)) = −(crs)
turnEntryLine � 2:
1+
In order to costhat
ensure γ p2p2.east
is always in the direction
= intercept.east + from p1 that forms a right-handed
10 · cos(crs)
t =tordistinguish between abs(crs) <= π and abs(crs) > π . (4.46)
fti
need
t�f i and tf i can be found
Intercept from
point, I α and αp respectively:
turnDist,
1 − cos γ p2.north = intercept.north
2 + 10 · 2sin(crs)
For abs(crs) <= π : 2
π/2 − χ
For�abs(crs) > π p2.east = intercept.east + 10 · cos(crs)
2:
where γ = 180◦ − ∆χ = 180◦ − (α + β)1 + cos γ p2.north = intercept.north + 10 · sin(crs)
tf i = r p2.east = intercept.east − 10(4.46)
· cos(crs)
α can be found from αp : 1 −
π/2 −cos
χ γ p2.north = intercept.north − 10 · sin(crs)
For abs(crs) > π2 :
L sin(αp ) p2.east = intercept.east − 10 · cos(crs)
where γ = 180◦ − ∆χ = 180◦ − (α + β) α = arcsin( Q
L�
) (4.47)
p2.north = intercept.north − 10 · sin(crs)
�
α can be found
where L� = from L2 +αpd:2 − 2Ld cos(αp )
L sin(αp )
β can be found from βp : α = arcsin( L�
) (4.47)
� LF I sin(βp )
Figure A-20: Calculation
where L� = L2 + d2 − 2Ld cos(αp ) α = arcsin(of turn� entry) line (4.48)
LF I
β can be found � �
where L�F I =fromL2FβIp :+ d2 − 2LF I d cos(β197 p ) with LF I = L2I + d2total − 2LI dtotal cos(135◦ and
L FI sin(β p )
βp = arcsin( LI sin(135
◦)
). α = arcsin( ) (4.48)
LF I L�F I
� � while first holding d otal and α constant
where L�F IThe
= result
L2F I +can
d2 be
− 2Lgraphed on d vs. E/W plots,
F I d cos(βp ) with LF I = L2I + d2total − 2LI dtotalt cos(135◦ and p
Figure 16: Calculation of turn entry line
and substituting
LI sin(135◦ ) a discrete set of L. This is illustrated in Figure 4-15.
 Ieast = Peast − t · sin(χ)

Inorth = Pnorth − t · cos(χ)

Since the slope of the line perpendicular to the current course is defined by m2 = − m11
and m = tan(α)

1
m1 = tan(90◦ − χ) =
tan(χ)
m2 = − tan(χ)

α = arctan(− tan(χ)) = −χ

In order to ensure that Q is always in the direction from I that forms a right-handed
π
coordinate system, need to distinguish between abs(χ) <= 2
and abs(χ) > π2 .

For abs(χ) <= π2 :

Qeast = Ieast + 10 · cos(χ)

Qnorth = Inorth + 10 · sin(χ)

For abs(crs) > π2 :

Qeast = Ieast − 10 · cos(χ)

Qnorth = Inorth − 10 · sin(χ)

198
A.4.3 lineCrossed

Inputs

P): current location

I: intercept of turn entry line with current straight line segment as returned by function
”getTurnEntryLine”

Q: second point on turn entry line returned by function ”getTurnEntryLine”

Inputs

b: returns -1 if turn entry line has been crossed at position P , 1 if the line is still ahead, 0
if P is on the turn entry line

Description

This function uses the vector cross product of vector V1 and V2 to determine whether the line
defined by points I and Q has been crossed. Figure A-21 illustrates the vector orientations.
The vector cross product is formed in a way, so that the resulting vector has negative
magnitude if P is behind the turn entry line (i.e. the turn entry line has been crossed) and
has positive magnitude if the turn entry line is still ahead.

−→
V1 = P I (A.51)
−→
V2 = IQ (A.52)

b = sign(V2 × V1 ) (A.53)

(A.54)

199
 intercept, I

vector, V_1 vector, V_2

aircraft
position, P turn entry line

current leg

Figure A-21: Geometry used to determine whether turn entry line has been crossed at
location, P

A.4.4 calcDistToGo

Inputs

• P: current aircraft position in local North-East coordinates

• Z: location of the point to which the distance to go should be estimated in local

North-East coordinates

• χ: course of the current leg

Outputs

R: distance to go from the current position P point Z

Trigger

Caller function (this is a library function to be used by several main modules)

200
Description

This function is used to calculate the ”Range-to-go-until the next segment”, R, which is used
by the Traffic Pattern Follower and the Energy Error Detector to time their switching to
the next leg or next segment of the traffic pattern respectively. In this case, the higher level
functions pass the current position, P and point Z as shown in Figure A-22. Furthermore,
this function is also used to determine the distance to go from the current position to the
next waypoint by the Traffic Pattern Follower and the Trajectory to Initial Point Follower.
In this case, the higher level functions simply pass the next leg fix, F in place of Z. Generally,
some small cross-track error c will always exist and hence the course connecting points P and
Z will not equal the desired course χ. Since the autopilot will attempt to zero the cross-track
error by returning the aircraft onto the line from point I to P prior to point Z, the range to
go is not simply the distance denoted by s. Rather, distance s denotes a lower bound on the
range to go until the start of the next segment. Distance c + d on the other hand denote an
upper bound to the range to go until the start of the next segment at point Z. In order to
arrive at a reasonable estimate of the range to go given the lack of knowledge of the exact
path that will be flown, distances s and c + d will be averaged. Note: It is necessary to
include the cross-track error c in the distance to go estimation since typical cross-track error
cause an elongation of the actual trajectory that causes energy errors on the order of the
errors that the energy management function is trying to fix. If, for example, c = 60 ft and
d = 1,500 ft, s differs by 4% from c + d. Using only s instead of the average of s and c + d
results in a difference of 2%, which is non-negligible considering that the energy manager
attempts to keep the energy error below 5%. The estimate for Range to Go from point P
until the start of the next segment at Z becomes:

s+c+d
R= (A.55)
2

All distances are calculated using the euclidean distance between the end points of their
connecting lines. Points P and Z are given as inputs. The intercept point, I can be calculated

201
 corrected on the base leg is determined by the difference in the length of trajectory:

1
∆E/W = (L� +L�F I −2t�f i +r·(180◦ −(α+β)))−(L+LF I −2tf i +r·(180◦ −(αp +βp )))
using library L/D
function ”calcInterceptPoint”.
av,tp0
(4.44)
L/Dav,tp0 can be written as:

L(L/Db ) + (LF I + LF )(L/D)f

L/Dav,tp0 = (4.45)
L + LF I + LF

t�f i and tf i can be found from α and αp respectively:

�
Start of turn segment, Z
1 + cos γ Next leg fix, F
tf i = r (4.46)
d 1 − cos γ t
Intercept, I
s
where γ = 180◦course
− ∆χ = 180◦ − (α
c + β)

α can be found from α p:

Position, P
L sin(αp )
α = arcsin( ) (4.47)
L�
Figure A-22:
� Calculation of Range-To-Go until start of the next segment
where L� = L2 + d2 − 2Ld cos(αp )
β can be found from βp :
LF I sin(βp )
α = arcsin( ) (4.48)
L�F I
�
� �
2
where LF I = LF I + d2 − 2LF I d cos(βp ) with LF I = L2I + d2total − 2LI dtotal cos(135◦ and
A.4.5 calcInterceptPoint ◦
βp = arcsin( LI sin(135
LF I
)
).
The result can be graphed on d vs. E/W plots, while first holding dt otal and αp constant
Inputs
and substituting a discrete set of L. This is illustrated in Figure 4-15.
For the second plot, L and dtotal are held constant and αp takes on a set of discrete values.
P: current position of the aircraft in the local North-East coordinate frame
This is illustrated in Figure 4-16.
Finally, L and αp are held constant while dtotal takes on a set of discrete values. Figure
P1, P2: North and East
4-17 illustrates this. coordinates of two points specifying the line to which the cross-
track error is to
For the be found;
trajectory for trajectory
updating legs,
strategy on the the
base leg,two
it is waypoints
possible that specifying the leg can
the course change
from the updated base leg to the final intercept leg is greater than 135◦ . This will cause
be used
136

Outputs

d: absolute value of the cross-track error

I: North-East coordinates of the intercept point, I of the line perpendicular to the line
defined by P 1 and P 2 through point P

202
Description

This function calculates the absolute value of the distance from point P to a line defined by
A Phase portrait for non-linear controller
P1 and P2. Figure A-23 illustrates the geometry of the problem. The line is extrapolated
In order to verify the performance and stability of the non-linear controller for any initial aircraft course and
position and any trajectory course and position, a phase portrait of course difference versus crosstrack error
past the two points and an has
intercept point,
been generated. The I is found
control at abycourse
law is given perpendicular
the following equations: to the course
k2 · e
χc = −arctan(
of the line. It is irrelevant whether an intercept in between the Vtwo ) (9)
T as points given is possible.
χ˙c = k1 (χc − χ) + χtraj
˙ (10)
The cross-track error, d is the distance from point P to intercept I. A is a point on the
where χc and χ represent the commanded and actual course of the aircraft resepectively, e represents the
crosstrack error, VT as the airspeed of the aircraft and χtraj
˙ the rate of change of the course of the trajectory.
line through p1 and p2 that khas
1 andthe
k2 aresame
tunableNorth
gains setcoordinate as P . 3Bandis1 awerepoint
to 3 and 1 respectively. chosenon
as athe
good line
compromise
between speed of convergence and potential to saturate the control output for small errors in crosstrack and
through P 1 and P 2 that has the
course same East coordinate as P . m1 , t1 , m2 , t2 designate slope
error.
The differential equations governing the rate of the change of the course difference of aircraft and trajec-
tory and the rate of change of the crosstrack error are as follows:
and intercept for the original and intercept line respectively. d and point I are found as
dχ k2 e
follows. dt
= k1 (−arctan(
VT as
) − χ) (11)
de
= Vg χ (12)
dt
Here, a straight trajectory is assumed or a trajectory with a constant rate of change. This allows dropping
the χtraj
˙ term. The resulting phase portrait is shown in Figure 23. All curves asymptotically converge to
the point (0,0) which indicates zero course difference and zero crosstrack error as desired.

psi´ = k1*(-1*atan(k2*(xtrk)/V)-psi) k1=3 k2=1

xtrk´ = V*psi V=12
10

xtrk
0

-2 Line
B
-4 P2
intercept, I
-6

-8

-10
-3 -2.5 -2 -1.5 -1 -0.5
P1 d v0 0.5 1 1.5 2 2.5 3
psi

Figure 23: Phase portrait of rate of change of course and

A h
31 position, P
aircraft

Figure A-23: Calculation of distance from line

203
 P 2north − P 1north
m1 =
P 2east − P 1east
t1 = P 1north − m1 · P 1east

Aeast = (Pnorth − t1 )

h = Peast − Aeast

φ = arctan(m1 )

d = h · sin(φ)
1
m2 = −
m1
t2 = Pnorth − m2 · Peast

Since the intercept point is on both the original and the perpendicular line:

t2 − t1
Ieast =
m1 + m11

Inorth = m1 · Ieast + t1

A.4.6 calcTurnDist

Inputs

χ1 : course of the current straight line segment to be flown in radians from −π to π.

χ2 : course of the next straight line segment to be flown in radians from −π to π.

r: turn radius to be assumed for the turn in m

204
Outputs

t: absolute value of the distance to the waypoint to start the turn at in m

Trigger

Function call

Description

This function calculates the distance to the waypoint to start the turn given the change
in course to be achieved. The geometry used for the calculation is shown in Figure A-24.
The calculation uses the fact that φ = 2τ . Applying the law of cosines to the two triangles
τ

τ φ

φ γ

d γ ∆χ
τ d
∆χ
s τ
φ τ
r
γ r
φ
φ

∆χ γ
γ
∆χ
∆χ
τ

∆χ
Figure A-24: Geometry used to determine the lead distance to the waypoint at which a turn
should be initiated

described by r and s and d and s to find an expression for s allows writing the following
equality:
r2 (1 − cos(π − γ)) = d2 (1 − cos γ) (A.56)

205
Ignoring the negative solution (since both d and r are positive quantities), yields the following
expression for the ratio of lead distance to turn d and turn radius r:
�
d 1 − cos γ
= (A.57)
r 1 + cos γ

where γ = π− | (χ1 − χ2 ) |.

206
Bibliography

[1] Technology review: Britain’s blind landing begins to pay off, New Scientist, 68 (1975),
p. 398.
[2] 2008 Nall Report - Accident Trends and Factors for 2007. http://www.aopa.org/asf/
publications/08nall.pdf, 2008.
[3] 2009 Nall Report - Accident Trends and Factors for 2008. http://www.aopa.org/asf/
publications/09nall.pdf, 2009.
[4] Wide-area augmentation system performance analysis report. http://www.nstb.tc.
faa.gov, October 2009.
[5] Aeronautical Radio, INC., Navigation Systems Data Base, ARINC Specfication
424-18, 2551 Riva Road, Annapolis, Maryland 21401-7435, 2005.
[6] AOPA, General aviation accidents 10-year trend. www.aopa.org/asf/publications/
topics/gatrend.pdf, 2005.
[7] E. M. Atkins, I. A. Portillo, and M. M. Strube, Emergency flight planning
applied to total loss of thrust, Journal of Aircraft, 43 (2006), pp. 1205–1216.
[8] Cirrus Design Corporation, Information Manual, Cirrus Design SR-22, 4515 Tay-
lor Circle Duluth, MN 55811.
[9] H. Diamond and F. W. Dunmore, A radio system for blind landing of aircraft in
fog, National Bureau of standards, pp. 678–685. Read before the Academy, September
19, 1930.
[10] H. L. Ehlers and J. W. Kraemer, Shuttle orbiter guidance system for the terminal
flight phase, Automatica, 13 (1977), pp. 11–21.
[11] FAA, Worldwide Soaring Turnpoint Exchange, Unofficial Coordinates of Soaring Con-
trol Points for North America. http://soaringweb.org/TP/NA.html.
[12] G. M. Hoffmann and C. J. Tomlin, Autonomous Automobile Trajectory Tracking
for Off-Road Driving: Controller Design, Experimental Validation and Racing, Pro-
ceedings of the American Control Conference 2007, pp. 2296–2301.

207
[13] D. Hughes, Ups, United complete GPS Autoland Tests, Aviation Week and Space
Technology, 141 (1994), p. 68.

[14] M. C. Jackson, T. M. Straube, T. J. Fill, T. M. Barrows, and S. M.

Nemeth, Onboard determination of vehicle glide capability for the shuttle abort flight
manager (safm), 2002.

[15] N. Meuleau, C. Plaunt, and D. E. Smith, Emergency landing planning for dam-
aged aircraft, 1989.

[16] R. Mola, History of aircraft landing aids. http://www.centennialofflight.gov/

essay/Government_Role/landing_nav/POL14.htm.

[17] J. Roskam, Airplane Flight Dynamics and Automatic Flight Controls, Darcorporation,
1995.

[18] J. Roskam and C. T. Lan, Airplane Aerodynamics and Performance, Darcorpora-

tion, 1988, ch. 8.

[19] S.-P. Shue and R. K. Agarwal, Design of Automatic Landing Systems Using Mixed
H2/H-Infinity Control, Journal of Guidance, Navigation and Dynamics, 22 (1999),
pp. 103–114.

[20] I. The MathWorks, Image processing toolbox. http://www.mathworks.com/

products/image/.

[21] T. Wagner and J. Valasek, Digital autoland control laws using quantitative feedback
theory and direct digital design, Journal of Guidance, Control and Dynamics, 30 (2007),
pp. 1399–1413.

[22] A. T. Wells and C. C. Rodrigues, Commercial Aviation Safety, McGrawHill,

2004, ch. 5.

[23] J. D. WILSON and T. K. Flesch, An Idealized Mean Wind Profile For The Atmo-
spheric Boundary Layer, Boundary-Layer Meteorology, (2003), pp. 281–299.

[24] J. A. Zak and W. G. Rodgers, Atmospheric Boundary Layer Wind Data during the
period Janunary 1, 1998 through January 31, 1999 at the Dallas-Fort Worth Airport,
Report CR-2000-210323, NASA, Aug. 2000.

208