An Optimal 3D Analytical Solution For Collision Avoidance Between Aircraft

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

An Optimal 3D Analytical Solution for

Collision Avoidance Between Aircraft


S. Luongo, C. Carbone, F. Corraro and U. Ciniglio
Flight Systems Department
CIRA (Italian Aerospace Research Centre)
Via Maiorise, 81043 Capua (CE), Italy
+39-0823-623330
s.luongo@cira.it

Abstract—This paper focuses on an optimal three- proved to be optimal with respect to the minimization of
dimensional analytical solution for aircraft non-cooperative aircraft deviation from its nominal trajectory. In particular,
collision avoidance. Based on a geometric approach, an an hypothesis to be considered is that aircraft speed vectors
analytical solution to a proper kinematic optimization keep constant (i.e., straight nominal trajectories): this is not
problem is here derived, which implies the simultaneous restricting if one considers a short-term time horizon and
change of all control variables (speed module, track and the solution is computed continuously, with a proper update
slope angles), thus this approach resulting very suitable for rate according with Nyquist-Shannon sampling theorem.
real-time applications. In a pair-wise non-cooperative Possible applications of this solution are its
collision avoidance, the speed vector of the aircraft implementation in an Autonomous Collision Avoidance
implementing the proposed control strategy is continuously (ACA) system to be installed onboard Unmanned Aerial
changed with the aim of skimming the safety bubble Vehicles (UAV) or simply as aid to pilots in finding the
surrounding the other aircraft (considered as an intruder).
right escape maneuver. Even if at present ACA does not
Under certain hypotheses, the proposed solution can be
distinguish between cooperative or non-cooperative aircraft
proved to be optimal with respect to the minimization of
conflict, in the future further research will be developed in
aircraft deviation from its nominal trajectory. Proper
performance indexes have been defined and challenging order to include sensors for the notification of the conflict
conflict scenarios, where the other aircraft be haves even a type and the algorithm will be accordingly updated.
pursuer, have been analyzed. 1 2 At present, all air carrier aircraft are equipped with
Traffic Alert and Collision Avoidance Systems, commonly
TABLE OF CONTENTS referred to as TCAS [2]. TCAS indicates the relative
altitude, distance, and bearing of transponder-equipped
1.INTRODUCTION ................................................................ 1 aircraft within a selected range, generally up to 40 miles; in
2.FORMULATION OF A KINEMATIC OPTIMIZATION addition to a traffic display, provides pilots with Resolution
PROBLEM FOR AIRCRAFT COLLISION AVOIDANCE ........... 2
Advisories when needed. The three main differences
3.ANALYTICAL RESOLUTION OF THE KINEMATIC
between ACA system (implementing the analytical solution
OPTIMIZATION PROBLEM ................................................... 3
proposed in this paper) and TCAS can be summarized as
4.NUMERICAL VALIDATION OF THE COLLISION
AVOIDANCE ALGORITHM ................................................... 5 follows:
5.CONCLUSION .................................................................... 7
REFERENCES ....................................................................... 8 (1) ACA system is completely autonomous, and can
BIOGRAPHY ......................................................................... 9 perform 3D collision avoidance manoeuvres, whereas
TCAS is designed only for vertical manoeuvres;
1. INTRODUCTION (2) ACA system can perform both cooperative and non-
cooperative collision avoidance manoeuvres (against
T HIS paper discusses an optimal three-dimensional
analytical solution for aircraft non-cooperative collision
avoidance. Based on a geometric approach, an analytical
non-equipped intruders), whereas TCAS works only in
a cooperative manner;
(3) ACA system performs optimal avoidance manoeuvres,
solution to a proper kinematic optimization problem is here i.e. the deviation from the nominal trajectory is
derived. This solution implies the simultaneous change of minimized.
all control variables: speed module, track and slope angles.
Motivated by the needs of designing an optimal 3D aircraft Geometric approaches used in literature lead to analytical
collision avoidance decision-making algorithm suitable for solutions for the Collision Avoidance problem in planar
real-time applications, this paper extends the results case only ([4]-[7]), whereas, when applied to a 3D
established in previous work [1]. environment, non-linear programming techniques are
Under certain hypotheses, the proposed solution can be usually required ([8]-[11]). Consider [1] for a review on
1
Collision Avoidance methods.
978-1-4244-2622-5/09/$25.00 ©2009 IEEE.
2
IEEEAC paper#1090, Version 3, Updated 2008:12:15 In paper [1] three different conflict resolution strategies

1
were investigated, each one involving a single control G
variable – χ (lateral-directional control), γ (longitudinal VB B

control) or V (speed control) – and a comparison among G G G


r = PB − PA
these strategies was carried out. Hence, the approach G
proposed in paper [1] results three-dimensional only for the VA

A
Conflict Detection process, while the Conflict Resolution, C

even though analytical, involves only one control variable at B


G
d AB
a time. In this paper, on the other hand, a real 3D analytical G G G
r = PB − PA
conflict resolution solution is designed, by extending the Plan

ξ
results of the previous work. At present, the algorithms is G G G
VAB = VA − VB
not developed to consider fixed constrains, as terrain, fixed A

obstacles and limited airspace. In this way, the development Fig. 1. Definition of minimum
G separation distance
of the decision-making algorithm is simplified, in order to vector d AB .
focus only the avoidance maneuvers with respect to the G
intruder aircraft. This means that the ACA algorithm, at Let d AB be a vector defined in [1] as the minimum
present development stage, is applicable in the case in separation distance experienced between aircraft, after a
which there are not constrains. In the future, in order to certain time horizon. It can be calculated as follows (see
complete the algorithm, it will be necessary to modify the Fig. 1):
algorithm proposed in this paper, taking into account fixed G G G
r ⋅V G G
constrains as terrain, fixed obstacles and limited airspace. d AB = G AB2 V AB − r . (1)
The effectiveness of the proposed solution is proved V AB
through Monte Carlo analysis, via intensive numerical
simulations, in conflict scenarios which take into account In paper [1] it has been proved that if a point object A
aircraft envelope limitation and dynamic constraints. Proper and a sphere B with radius R are moving in a 3D
G
performance indexes have been defined and challenging environment with constant velocity vectors, respectively V A
conflict scenarios, where the other aircraft be haves even a G
and V B , they are headed for a collision if and only if the
pursuer, have been analyzed.
following conditions are satisfied (Conflict Detection
conditions):
2. FORMULATION OF A KINEMATIC
OPTIMIZATION PROBLEM FOR AIRCRAFT G
d AB < R and r < 0. (2)
COLLISION AVOIDANCE
According to the collision geometry proposed in [1],
The Conflict Resolution strategy proposed in this paper is
consider the aircraft with ACA module on-board (A/CA)
based on the analytical solution of the following kinematic
modeled as a point object – with 3 Degree Of Freedom
G optimization problem: find the minimum change in nominal
(DOF) – having velocity V A , while the other aircraft (A/CB, trajectory of aircraft A/CA to be forced (compatible with its
considered as an intruder) modeled as a sphere with radius envelope limitation and dynamic constraints) in order to
G
R (safety bubble) having velocity VB . The relative velocity avoid a collision with the safety bubble surrounding aircraft
G G G A/CB.
vector V AB = V A − VB transforms the kinematic collision G G
avoidance problem – where both aircraft are moving – into Let PA (t ) = [ x A (t ), y A (t ), z A (t )]T and PB (t ) be,
G
an equivalent problem where aircraft A/CB is stationary (the respectively, A/CA and A/CB nominal trajectories. PA (t )
sphere) and point object A/CA moves with relative velocity G
G can be expressed in terms of speed vector V A (t ) along the
V AB with respect to it (see Fig. 1) [7].
nominal trajectory as follows
G G t G
PA (t ) = PA (t0 ) + ∫ VA (τ )dτ . (3)
t0

G
Let PAd (t ) be the modified trajectory resulting from the
demanded speed vector function
K K
( )
V Ad (t ) ≡ V Ad V Ad , χ Ad , γ Ad , t . The deviation from the
nominal trajectory of aircraft A/CA is

2
G G t G G
[
PAd (t ) − PA (t ) = ∫ VAd (τ ) − VA (τ ) dτ . ] (4) G t G G
∫ ΔVA (τ )dτ = ΔVA (t0 ) (t − t0 )
t0
ΔPA (t) = (6)
t0
Minimizing nominal trajectory deviation – under subject to constraint (5)-1, which can be rewritten as
envelope limitation and dynamic constraints – means follows:
G G G G
r (t ) + ΔV A (t0 )(t − t0 ) ≥ R, ∀t ≥ t0 .
t
minimizing the quantity ΔPA (t) = ∫
t0
ΔVA (τ )dτ as stated (7)

by the following nonlinear programming problem:


Hence, the minimization of A/CA nominal trajectory
G
t G deviation is equivalent to the minimization of vector ΔV A ,
min
VAd ,χ Ad ,γ Ad ∫
t0
ΔVA (τ )dτ
subject to Collision Avoidance constraint (7).
G G Assumption (ii) implies that
⎧ PAd (t) − PB (t) ≥ R, ∀t ≥ t0 G G G G G G G Gd G
⎪⎪
s.t. ⎨VAd (t) ∈[VAmin,VAmax], γ Ad (t) ∈[γ Amin,γ Amax]
(5) (
Δ V A = V Ad − V A = V Ad − V B − V A − V B = V AB
G
) (
− V AB , )
G G
⎪ and r (t ) = r (t 0 ) + V AB (t − t 0 ) . Thus, constraint (7) can be
⎪⎩ ∃taA > 0.
rewritten as
G Gd
Constraint (5)-1,
r (t0 ) + VAB (t − t0 ) ≥ R, ∀t ≥ t0 (8)
G G G t G
PAd (t) − PB (t) = r (t) + ∫ ΔVA (τ )dτ ≥ R G Gd
t0 It can be proved that at time instant t = t * ≡ r (t 0 ) ⋅ V AB ,
Gd 2
ensures that minimum separation distance R is never V AB
violated (Collision Avoidance); constraint (5)-2 represents G Gd
the aircraft envelope limitations; (5)-3 is a dynamic quantity r (t0 ) + VAB (t − t0 ) is minimum and results equal
constraint, since the closed-loop system “Aircraft + G
d d
G
to d AB . Consequently, by imposing d AB ≡ R (tangency
Autopilot” has a finite settling time t aA for speed vector
G with the safety bubble), constraint (7) is certainly satisfied.
changes, i.e., V Ad cannot be reached instantaneously but
Finally, general problem (5), under assumptions (i)-(iii),
requires a certain time t aA . becomes
G
min ΔVA
VAd , χ Ad ,γ Ad
3. ANALYTICAL RESOLUTION OF THE Gd (9)
KINEMATIC OPTIMIZATION PROBLEM s.t. d AB = R.

A. Introduction
This new problem admits an analytical solution, as it will
In order to approach analytically the general Collision be proved in the next paragraph.
Avoidance problem formulated in the previous paragraph,
some simplifying assumptions are here required. Thanks to
these assumptions an analytical solution will be derived in B. Geometric Formalization
G
§III-B, representing a global optimum to optimization It is required to find the minimum ΔV A to be applied to
problem (5). G
Afterwards, in §III-C and §III-D, it will be shown how to speed vector V A such that the new speed vector
G G G
remove these assumptions in order to take into account V Ad = V A + Δ V A results tangential to the safety bubble.
envelope limitation and dynamic constraints expressed in Kd
Let Vˆ d be the unit vector of V AB . In the plane formed
problem (5). AB
G
by unit vectors VˆAB and VˆAB
d
, the minimum ΔV A can be
Three assumptions are hereafter considered: G
i. Change in speed vector occurs only at time t0, i.e., obtained from the orthogonal projection of V AB on
G
ΔV A (t ) is a step function [4]. direction vector VˆAB
d
, (cf. Fig. 2): the module of any other
G G
ii. Straight aircraft nominal trajectories at constant vector Δ V A′ is greater than ΔV A . Hence,
speed modules. G
iii. No aircraft envelope limitations and dynamic ΔVA = VAB sinΩ , with Ω ∈ [0, π2 [ .
constraints.
Thanks to assumptions (i)-(iii), general problem (5) can
be reduced to minimize the following function

3
G
A VAB X
VˆAB
d
Ω Plane π C
G
ΔVA G K B
r
ΔVA′ G
d AB
Fig. 2. Plane formed by VˆAB and VˆAB
d
unit vectors. r̂G Gd
η
G θ ξ VAB d AB
In order to compute minimum ΔV A , direction VˆAB – Gd
d
VAB
which guarantees tangency with the safety bubble and is
A
characterized by a minimum angle Ω – has to be Y
computed. At this aim, consider circumference C⊥ (cf. Fig. r̂ ⊥
Gd
3), obtained from the intersection between the cone with a Fig. 4. V AB vector computation on plane π.
R G
generatrix at angle η = sin−1 G to axis r , and the safety
r Angle Ω = η − ξ is minimum when η has the same
bubble. sign of ξ . Two cases can be considered: ξ ≠ 0 and
ξ = 0.

1) Case ξ ≠0 (general case)


d
K
From Fig. 4, the expression of V AB which minimizes
C⊥ R ⊥ B R G
ΔV A can be derived as follows (some trivial manipulations
B' have been here omitted for the sake of brevity)
G
r D Gd Gd G
η ξ C VAB d
= VA −VB = VAB sinη ⋅ rˆ⊥ +VAB
d
cosη ⋅ rˆ =

G G G
C*
=
VAB cos(η −ξ )
sinξ
[
sinη ⋅VˆAB − sin(η −ξ ) ⋅ rˆ . ] (10)

A V AB = V A − VB Plane π

Notice that expression (10), under assumptions (i)-(iii), is a


C⊥ global optimum to the general problem (5).
B' R⊥
2) Case ξ = 0 (singular case)
This singularity is due to the alignment between vectors
D G G
V AB and r . In this case, plane π is not univocally defined
C but it has to be chosen in the sheaf of planes having as axis
C* G
the direction of vector r .
Fig. 3. Definition of circumference C⊥.

For each point C belonging to circumference C⊥, direction C. Aircraft Envelope Limitation and Dynamic Constraints

AC identifies a possible unit vector VˆAB


d
tangent to the Solutions γ Ad and V Ad , which can be derived from (10)
safety bubble. However, point C* – lying on the plane obtained from purely geometric considerations, do not take
G G into account aircraft envelope limitation constraints, which
formed by V AB e r (plane π) – is the only one which
G can be expressed as γ Ad ∈ [γ A min , γ A max ] and
ensures the minimum Ω (and hence the minimum ΔV A )
V Ad ∈ [V A min , V A max ] (as for χ A , there are no limits). If
d

since DC * ≤ DC for all C belonging to circumference C⊥. γ Ad ∉ [γ A min , γ A max ] and/or V Ad ∉ [V A min , V A max ] a control
Kd
Hence, an analytical expression for V AB has to be searched
strategy could be to force the envelope limit for γ d
A and/or
in plane π, as shown in Fig. 4.
V Ad , and search for a χ Ad which ensures the tangency with

4
the safety bubble surrounding aircraft A/CB, as explained in bubble with radius R ' . In this case, solution (10) cannot be
[1]. adopted since it does not exist a tangent to a sphere of
Aircraft A/CB (intruder) is assumed to be surrounded by a radius R ' passing through a point inside the sphere.
virtual safety bubble, which defines a forbidden zone for Hence, while A/CA lies within the safety-bubble with
A/CA whose maneuvers aim to do not enter the safety radius R ' , an alternative strategy is required; for this
G
bubble. Solution (10) is optimal with respect the deviation situation, the demanded speed vector V Ad ,e to be assigned to
from the nominal trajectory only under the assumption of the autopilot of aircraft A/CA may be defined as in Fig. 5.
straight trajectories and single step change in velocity The adopted strategy provides a “strong” control action
vector. However, it has to be considered that the closed- aiming at a quick aircraft exit from the safety bubble of
loop system “Aircraft + Autopilot” has a certain settling radius R ' , after that the tangential solution (10) will
time t aA for speed vector changes, i.e., A/CA speed vector continue to be applied.
cannot be changed instantaneously but requires a certain
time t aA .
If the intruder moves straight and with constant speed
module, this is not a real problem provided that initial
conflict detection happens at a range long enough
(depending on aircraft performances); in fact, conflict
detection conditions (2) can be evaluated continuously, with
proper update rate 1/TS, according to the Nyquist-Shannon
sampling theorem and, consequently, reference signals to
the autopilot are generated with sampling time TS. On the
other hand, if the intruder is non-cooperative (or, even
worse, a pursuer) there is no guarantee that A/CA does not
enter the safety bubble surrounding A/CB.
In order to take into account this dynamic constrain,
safety bubble radius R may be designed coherently. In
particular, in this paper it has been chosen to enlarge the
safety bubble as R ' = R + ΔR , where:
Fig. 5. Escape trajectory for the exit from safety bubble
• R = 500 ft ≈ 152.4 m (derived from the FAA incident of radius R’.
criteria, see also [3]) represents the minimum allowed
separation distance, under which a collision happens; 4. NUMERICAL VALIDATION OF THE COLLISION
• ΔR can be designed assuming a conservative AVOIDANCE ALGORITHM
estimation for all possible maneuvers that the intruder
The effectiveness of the proposed approach has been
can perform during time t aA : ΔR = t aA ⋅ VB MAX . proved through Monte Carlo analysis, via intensive
If, for instance, t aA ≈ 10s and VB MAX ≈ 45 m / s , then numerical simulations, carried out in proper 3D collision
scenarios, as shown in Fig. 6. These test scenarios have
R ' = R + ΔR ≈ 4 R = 2000 ft . Simulation results of been designed according to standard specification given in
paragraph 4 will show that, provided that A/CB has the same [12].
performances of A/CA, even though A/CB is a pursuer, a Conditions (2) also offer an analytical criterion to find
safety bubble radius R ' ≈ 4 R is sufficient to assure the initial aircraft velocity vectors such that a collision situation
collision avoidance (i.e., the relative distance between is certain to happen, so to drastically reduce the amount of
aircraft is always greater than R = 500 ft ). collision geometries to be analyzed, as explained in [1].
In each conflict scenario it is assumed that intruder speed
vector is initially constant and such that a collision will
D. Emergency: Safety Bubble Exit Strategy certainly happen: Fig. 6 shows – for a given intruder (A/CB)
It can be proved that a safety-bubble radius R ' = R + ΔR , initial position – a “cone” of speed vectors that guarantee a
with ΔR properly designed, guarantees that the minimum collision with A/CA, if a collision avoidance maneuver is
distance between aircraft is always greater than R (the not performed. A/CA is of class VLA and modeled by a
analytical proof is here omitted for the sake of brevity) – detailed 3DOF aircraft model having VA0 = 45 m/s, VA
provided that A/CB has the same performances of A/CA and ∈[25, 55] m/s and γA ∈[-8°, 8°].
the initial relative distance is greater than a certain critical The proposed scenario foresees about 5000 starting
distance – even though A/CB is a pursuer. Nevertheless, for
points (r,Ψ,Θ) for the intruder, randomly chosen, with
some scenarios it could happen that A/CA enters the safety-

5
uniform distribution: Performance indexes S and D have been computed in
r = 2.5 n mi, Ψ ~ U[-110°, 110°], Θ ~ U[-15°, 15°]. each scenario. Fig. 8 shows minimum separation distance S,
which is always within a range of ± 3 cm (|S| ≤ 0.03, i.e. no
collisions) and very close to the origin, which means that all
avoidance maneuvers skim the safety bubble. This control
strategy, besides avoiding collisions in the proposed
scenario, offers the advantage of reducing the maximum
deviation D from the nominal trajectory, as shown in Fig. 9,
where Cumulative Distribution Function (CDF) associated
with random variable D (maximum deviation from the
nominal trajectory) is represented.
Fig. 10, Fig. 11 and Fig. 12 disclose that initial velocity
vector changes (in terms of module, track and slope angles)
are very small.
Fig. 6. 3D test scenarios definition.

Two performance indexes have been defined as follows


(Fig. 7):
S – Minimum separation distance between aircraft
A/CA trajectory and the safety bubble of aircraft A/CB. S<0
means collision.
D – Maximum deviation of aircraft A/CA from its
nominal trajectory, representing the distance to cover in
order to return on the nominal trajectory, as a consequence
of the avoidance maneuver.

Fig. 8. Scenario 1: minimum separation distance S.

0.9

0.8

0.7
CDF: Pr(D ≤ d)

0.6

0.5

0.4

0.3

Fig. 7. Performance indexes S and D. 0.2

0.1
Hereafter two kinds of scenario are considered, in order
to test the effectiveness of the proposed approach both for 0
20 40 60 80 100 120 140 160
nominal (the intruder moves with stationary speed vector) d [m ]
and critical situations (the intruder is a pursuer). Fig. 9. CDF of random variable D: μ = 52.487 m,
A. Scenario 1: Intruder moves with stationary speed vector σ = 37.975 m, Pr(D ≤ 125.89 m) = 0.95, Pr(D ≤ 148.94 m)
= 0.99, min(D) = 0.0007 m, max(D) = 177.3 m.
Simulations have been carried out under the following
hypotheses:
A-i. Safety bubble radius: R’=R = 500 ft.
A-ii. A/CB stationary speed module is randomly
chosen: VB ~ U[45, 90] m/s.

6
1 B. Scenario 2: Pursuer intruder
0.9 For this kind of scenario instead, simulations have been
0.8 carried out under the following hypotheses:
B-i. Safety bubble radius: R' = R + ΔR ≈ 2000 ft.
CDF: Pr(|ΔV| ≤ |δ V|)

0.7

0.6
B-ii. A/CB has the same performances of A/CA and its
initial speed module is randomly chosen:
0.5
VB ~ U[25, 55] m/s.
0.4 B-iii. A/CB is equipped with a pursuit algorithm
0.3 implementing (10), where the role of A/CA and
0.2 A/CB is exchanged and the safety bubble is
0.1
reduced to a point.
Also for this more challenging scenario there are no
0
0.5 1 1.5 2 2.5 collisions, since minimum separation distance S is always
|δ V| [m /s]
positive (cf. Fig. 13). In Fig. 14, a time history of relative
Fig. 10. CDF of random variable | δV|: μ = 0.317 m/s, distance r(t) between aircraft is shown for 100 scenarios. It
σ = 0.368 m/s, Pr(|δV| ≤ 1.11 m/s) = 0.95, Pr(|δV| ≤ 1.61 is interesting to notice two facts:
m/s) = 0.99, min(|δV|) = 3.45e-6 m/s, max(|δV|) = 2.83
m/s.
1. There are few scenarios where A/CA enters the
1 safety-bubble with radius R' ( r (t ) < R' ) and strategy
0.9 of paragraph 3.D is required, acting successfully by
converging towards a tangential condition with the
0.8
sphere of radius R' .
CDF: Pr(|Δχ| ≤ |δχ|)

0.7
2. All scenarios converge towards an equilibrium
0.6 situation where both aircraft move parallel (e.g., Fig.
0.5 15).
0.4

0.3

0.2

0.1

0
1 2 3 4 5
|δ χ | [d eg ]
Fig. 11. CDF of random variable |δχ|: μ = 0.769°,
σ = 0.808°, Pr(|δχ| ≤ 2.46°) = 0.95, Pr(|δχ| ≤ 3.57°) = 0.99,
min(|δ χ|) = 8.69e-6°, max(|δχ|) = 5.582°.

0.9

0.8
CDF: Pr(|Δγ| ≤ |δγ|)

0.7

0.6
Fig. 13. Scenario 2: minimum separation distance S.
0.5

0.4 5. CONCLUSION
0.3 In this paper an optimal three-dimensional analytical
0.2 solution for aircraft collision avoidance has been proposed.
0.1 Based on a geometric approach, an analytical solution to a
proper kinematic optimization problem has been derived,
0
0 1 2 3 4 5 involving the simultaneous change of all control variables
|δ γ | [d eg ]
(speed module, track and slope angles), thus this approach
Fig. 12. CDF of random variable |δγ|: μ = 0.944°,
resulting very suitable for real-time applications.
σ = 0.892°, Pr(|δγ| ≤ 2.83°) = 0.95, Pr(|δγ| ≤ 3.81°) = 0.99,
In order to test the effectiveness of the proposed
min(|δγ|) = 0°, max(|δγ|) = 5.15°.
approach, two kinds of scenario have been considered, both
for nominal (the intruder moves straight and with constant

7
speed module) and critical situations (the intruder is a latter kind of scenario (more challenging) also there have
pursuer having the same performances of the evader). been no collisions. Moreover, simulation results have
shown two interesting facts:

• There are just few scenarios where aircraft A/CA enters


the enlarged safety-bubble (to take into account aircraft
dynamic constraints) having radius R'≈ 4 R (about
6%).
• All scenarios converge towards an equilibrium situation
where both aircraft move parallel.

Real-time hardware in the loop simulations and flight


tests on a Very Light Aircraft (VLA) flying test bed
(FLARE – Flying Laboratory for Aeronautical REsearch)
have been scheduled and are currently in progress (Fig. 16).
Further research concerns the possibility of extending this
approach to the case of multiple cooperative/non-
cooperative aircraft and with the ellipsoidal safety bubble.
Fig. 14. Time history of relative distance r(t) between
Moreover, will be developed in order to include sensors for
aircraft for a sample of 100 scenarios.
the notification of the conflict type (cooperative or non-
cooperative) and the algorithm will be accordingly updated.
In the future, in order to complete the algorithm, it will be
necessary to modify the algorithm proposed in this paper,
taking into account fixed constrains as terrain, fixed
obstacles and limited airspace.

REFERENCES
[1] C. Carbone, U. Ciniglio, F. Corraro, S. Luongo, “A Novel 3D
Geometric Algorithm for Aircraft Autonomous Collision Avoidance”,
45th IEEE Conference on Decision and Control (CDC’06), San
Diego, CA, USA, December 2006.
[2] “Introduction to TCAS II, Version 7”, U.S. Department of
Transportation, Federal Aviation Administration, November 2000.
[3] NASA ERAST, “Non-Cooperative Detect, See, & Avoid (DSA)
Sensor Study”, July 2002.
[4] K.-D. Bilimoria, “A Geometric Optimization Approach to Aircraft
Conflict Resolution”, AIAA Guidance, Navigation, and Control
Conference, August 2000, Denver, Colorado.
[5] P.A. Wilson and C.J. Harris, “A Line of Sight Counteraction
Fig. 15. Pursuit-evader scenario example. Navigation Algorithm for Ship Encounter Collision Avoidance”, The
Journal of Navigation (2003), No. 56, pp. 111-121.
[6] A. Chakravarthy and D. Ghose, “Obstacle Avoidance in a Dynamic
Environment: A Collision Cone Approach”, IEEE Transactions on
Systems, Man and Cybernetics – Part A: Systems and Humans, Vol.
28, No. 5, September 1998.
[7] P. Fiorini and Z. Shiller, “Motion Planning in Dynamic Environments
using Velocity Obstacles”, Int. Journal of Robotics Research, Vol.17,
No.7, pp.760–772, July 1998.
[8] J. Gross, R. Rajvanshi, and K. Subbarao, “Aircraft Conflict Detection
and Resolution using Mixed Geometric and Collision Cone
Approaches”, AIAA Guidance, Navigation, and Control Conference
and Exhibit, Providence, Rhode Island, 2004.
[9] M.-A. Christodoulou and S.-G. Kodaxakis, “Automatic Commercial
Aircraft-Collision Avoidance in Free Flight: The Three-Dimensional
Problem”, IEEE Trans. on Intelligent Transportation Systems, Vol. 7,
No. 2, June 2006.
[10] J. Hu, M. Prandini, and S. Sastry, “Optimal Coordinated Maneuvers
Fig. 16. CIRA flying laboratory. for Three-Dimensional Aircraft Conflict Resolution”, AIAA Journal
of Guidance, Control, and Dynamics, Vol.25, No.5, pp. 888-900,
In the first kind of scenario, this control strategy, besides September-October 2002.
[11] A.U. Raghunathan, V. Gopal, D. Subramanian, L.T. Biegler, and T.
avoiding all collisions, has shown the advantage of reducing Samad, “Dynamic Optimization Strategies for Three-Dimensional
the maximum deviation from the nominal trajectory. In the Conflict Resolution of Multiple Aircraft”, AIAA Journal of Guidance,

8
Control, and Dynamics, Vol.27, No.4, pp. 586-594, July-August design, TLT/BOEING-IWT (Icing Wind Tunnel) Spray-bar
2004.
auxiliaries dynamic modelling and control, ASICA (Air
[12] “Standard Specification for Design and Performance of an Airborne
Sense-and-Avoid System (F2411-04)”, ASTM International, 2004. Management Simulation for Aircraft Cabin) of the 5th
Framework Programme, GARTEUR AG-11 New Analysis
BIOGRAPHY techniques for Clearance of Flight Control Laws.

Salvatore Luongo received the Umberto Ciniglio graduated cum laude


Laurea degree (cum laude) in with degree in Electronics
Aerospace Engineering from the Engineering in 1992. Research
University of Naples “Federico II”, Engineer in the Flight System
Italy, in 2004. He is currently Department at CIRA since 1992, was
Researcher at the Flight Systems involved in activities mainly concerning
Department of CIRA (Italian modelling and control of aircraft both
Aerospace Research Centre), Capua, in the civil and military field. In
Italy. Participation in IFATS project particular he worked on the application
(Innovative Future Air Transport System) funded by EC of modern multivariable techniques to flight control law
under the VI FP. His research interests are in modeling and design, and developed some specific experiences in the field
control aerospace systems and decision-making algorithms of digital flight control system validation, by means of both
with applications to air traffic and autonomous collision real-time hw in the loop simulations and experimental flight
avoidance systems. test by using small scale unmanned flying platforms. He
participated as technical responsible of CIRA activities in
Ciro Carbone received the Laurea several European research projects both within the
degree (cum laude) in Electronic GARTEUR organization (FM AG-08 from 93 to 95), both in
Engineering from the University of the IV and V frameworks (ADFCS & ADFCSII from 1997 to
Naples “Federico II”, Italy, in 2001 2003).
and the Ph.D. degree in computer
engineering from the University of
Salerno, Italy, in 2005. He was
Invited Scientist at the Siemens AG
Corporate Technology, Software &
Engineering Division, Munich,
Germany, in 2001, and Visiting Researcher at the SYSTeMS
group of the University of Gent, Belgium, in 2004.
Participation in IFATS project (Innovative Future Air
Transport System) funded by EC under the VI FP.

Federico Corraro graduated cum


laude with degree in Electronics
Engineering in 1992. Research
Engineer in the Flight System
Department at CIRA since 1994,
was involved in activities mainly
concerning modelling and control
of aircraft both in the civil and
military field. In particular he
worked on the application of modern multivariable
techniques to flight control law design, and developed some
specific experiences in the field of digital flight control
system validation, by means of both real-time hw in the loop
simulations and experimental flight test by using small scale
unmanned flying platforms. He was on contract professor of
Control System Technologies, at the Department of
Automation Engineering of University of Naples “Federico
II” (from 2001 to 2004). He participated as technical
responsible of CIRA activities in several international
research projects: ESA-PWT (Plasma Wind Tunnel)
Dynamic Modelling and Control, ASE/TLT-LSWT (Low
Speed Wind Tunnel) Automation and Control System

You might also like