Professional Documents
Culture Documents
A Mathematical Model of A Twin Ducted Fan VTOL Jetpack
A Mathematical Model of A Twin Ducted Fan VTOL Jetpack
A Mathematical Model of A Twin Ducted Fan VTOL Jetpack
net/publication/260795523
Article in Proceedings of the Institution of Mechanical Engineers Part G Journal of Aerospace Engineering · January 2015
DOI: 10.1177/0954410013503060
CITATIONS READS
4 5,626
3 authors, including:
Mathieu Sellier
University of Canterbury
167 PUBLICATIONS 1,628 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mathieu Sellier on 05 February 2015.
Michael Speck
*
Department of Mechanical Engineering, University of Canterbury, Christchurch, New Zealand
Jörg Buchholz
Department of Mechanical Engineering, Hoschschule Bremen, Germany
Mathieu Sellier*
March 2013
Abstract
A dynamic model of a twin ducted-fan VTOL (vertical take-off and landing) aircraft, the Mar-
tin Jetpack, has been developed to study and improve the understanding of the flight mechanics
involved with this novel aircraft concept. This paper describes the flight mechanics of a twin
ducted-fan aircraft and explains in detail the modeling of the forces and moments contributed by
the, twin ducted-fans, body aerodynamics, control surfaces, gyration, and landing gear interac-
tions. Also, a novel model for the movement of the duct center of pressure has been developed,
which allows for the complex duct pitching moment to be predicted. Employing the conventional
aircraft modeling methodology, a system of ordinary differential equations (ODEs) that describes
the behaviour of the aircraft is developed. The equations are solved in real-time using Matlab-
Simulink software to simulate the response to given inputs. Comparison of the flight data with
both steady state (trimmed) and dynamic simulations show good agreement, which validates the
novel duct centre of pressure model. The validated model allows the aircraft designer/engineer to
efficiently evaluate the sizing of key aerodynamic features and various control methodologies to
1
1 Introduction
The Martin Jetpack, Figure 1, is a new vertical takeoff and landing (VTOL) aircraft concept that is
being developed as a solution for personal flight. This aircraft, like other VTOL aircraft, produces
a lift force greater than the weight force of the aircraft at zero airspeed allowing it to hover. The
Martin Jetpack achieves the required lift force for flight utilizing two ducted-fans in twin configuration
powered by a single two-stroke combustion engine. Due to the low inertia of the Jetpack the rate of
divergence is very high, so the Jetpack features a fly-by-wire system to introduce stability and decrease
the pilot workload. Control about the three axes longitudinal, lateral and vertical axis (x, y, and z
axis respectively, refer to Figure 1) is achieved by roll, pitch, and yaw vanes. These vanes are immersed
in the high dynamic pressure slipstreams of the ducted-fans allowing the aircraft to maintain control
independent of the aircraft’s airspeed. A fourth control is used to control the rate of climb by varying
engine power/speed, and hence, fan speed and thrust. The Martin Jetpack exploits the ducted-fan
propulsion system, which allows for a compact form (height 1650 mm, width 1710 mm, and length
1520 mm), a maximum takeoff of weight of 250 kg, and a target cruise speed of 100 km/h.
A ducted-fan has numerous aerodynamic performance benefits compared to an open propeller that
make it appealing for VTOL aircraft manufactures. The addition of the duct around the propeller
allows for the fan blades to maintain loading out to the blade tip [1]. The duct also defines the
final flow expansion and prevents the flow contraction that would otherwise occur behind an open
propeller. The benefit of the flow expansion is that a higher thrust, particularly at static conditions,
can be obtained for an equally sized propeller for the same input power. The larger the expansion
ratio the greater the static thrust improvement over an open propeller. However, an upper limit exists
on the expansion ratio as too high a value leads to flow separation within the duct rendering further
expansion ineffective. A theoretical comparison between an open propeller and a ducted-fan can be
done, as by [2], by taking the ratio of ducted-fan and open propeller static power equations as:
q
Td3 23 12
Pd 4ρAr d 1 Td Ap
= q 3 =√ (1)
Pp Tp 2d Tp Ar
2ρAp
where the expansion ratio d is defined by the ratio of duct exit to fan rotor areas as:
2
Radiator ducting
Ducted-fan
Engine
v,y
q
Pilot
Pitch/roll/yaw
joystick Yaw vane
ζ
Pilot display
Throttle
joystick
p Roll vane
u,x
Fuel tank ξ η Pitch vane
Pogo, main
r landing gear
w,z
Figure 1: The Martin Jetpack P-11A sign convention and key features
3
Ad
d = (2)
Ar
Using equation (1) it can be seen that for a ducted-fan with an expansion ratio of εd = 1, and
sharing the same diameter as an open propeller and the same power the ducted-fan will produce 26 %
more thrust, or for the same thrust will require only 71 % of the power of an open propeller, which is
similar to the result shown in [1]. An experimental investigation by [2] who measured the effect of: inlet
radius, expansion ratio, duct chord length, blade tip clearance and blade collective angle, found that
the addition of the duct could provide up to a 94 % increase in thrust over an open propeller sharing
the same diameter and power, and that a duct diffuser angle of 10◦ was found to be the optimum.
Numerous investigations into the effects of duct inlet geometry on the thrust and control have been
made [3, 4, 5]. Increasing the inlet or duct lip radius has been shown to improve the static thrust of a
ducted-fan. This increase in thrust is due to the reduced restriction on the flow entering the duct with
increased radius, which allows for a greater mass flow through the duct, and hence, greater thrust.
The increase in thrust manifests as a lower static pressure over the inlet radius, where increasing the
inlet radius increases the area that the lower static pressure acts upon. This correlates well to the
bell mouthed shaped inlets used for industrial piping and ducting applications [6]. The inlet radius
of curvature also plays an important part on the pitching moment of a ducted-fan in crosswind flight.
Smaller, sharper inlet radii, encourage the flow to separate early, which can dramatically change the
nature of pitching moment. Larger, rounder inlet radii have the opposite effect, [7].
The presence of the duct improves the lift generating capability of the fan blades as the duct
effectively prevents tip vortices from forming, providing the tip clearance is kept small, less than 1 %
of the rotor radius [6]. Stator blades can be employed in ducted-fans to improve thrust by recovering
the induced swirl flow, and to create a counter acting torque that balances the fan torque resulting in
a torque neutral ducted-fan, as used on the Martin Jetpack. The duct also has the practical advantage
of shielding the spinning fan, which makes it inherently safer than an open propeller when in close
proximity.
However, the presence of the duct adds to the overall weight of aircraft and also becomes detrimental
at higher airspeeds where the duct drag exceeds the thrust benefits of a duct. The ducted-fan is best
4
used for applications in low speed flight where the propeller diameter is limited and/or needs to be
shielded for functionality. Hence, the Martin Jetpack exploits the advantages of ducted-fan system.
The original work on ducted-fan technology originated in the post war era of the 1950s. NASA
performed many wind-tunnel experiments [8, 9, 10] on wing-tip ducted-fan aircraft configurations, such
as the VTOL Doak-16 [11]. The Hiller flying platform VZ-1 Pawnee [12] consisted of a single large
duct with counter rotating rotors, where pilots used kinaesthetic movement of their bodies to control
the aircraft. However, both the Doak-16 and VZ-1 as well as numerous modern unmanned ducted-fan
aircraft have had trouble flying at high speeds or in gusty conditions [13] due to large uncontrollable
The reason being that ducted-fans create a large aerodynamic force normal to the duct axis, as
per [14], due to the momentum change of the air flow through the duct, which will be referred to
as the ducted-fan normal force. The location of the ducted-fan normal force, which is effectively the
ducted-fan centre of pressure, is a function of: airspeed, angle of attack, mass flow rate through the
duct, and duct geometric properties – and changes significantly as flow conditions change. To avoid the
unnecessarily large pitching moments from the duct, as reinforced by [4], the designer of a ducted-fan
aircraft needs to best align the duct centre of pressure (CP) to the centre of gravity (CG).
The state-of-the-art method for controlling ducted-fan VTOL aircraft has been through use of
control vanes located in the wake of the ducts, where 0.75 to 1.0 rotor diameters below the duct exit
plane has been found to be the optimum placement for control vanes [5]. This method of control
has proven to be the simplest and most effective to implement. However, other methods such as
synthetic jets [15] and ejectors [16] have shown potential for producing strong control moments, which
An accurate mathematical model of an aircraft allows for a higher understanding of the flight
behaviour, which leads to a reduced number of test flights and consequently, reduced cost and risk.
An accurate model allows designers to investigate the effect of changing model parameters of not just
the physical aircraft, but also those of the control systems included in the model [17]. For unique
flying machines which have unfamiliar characteristics it is prudent to have real-time simulation of the
5
This paper describes the development of a mathematical model that describes the unique dynamic
behaviour of a twin ducted-fan VTOL aircraft, namely the Martin Jetpack. Hence, the novelty of
this paper is in the application of the conventional aircraft modeling methodology to the unique
configuration of Martin Jetpack. An original model for the determination of the ducted-fan moment
as a function of the angle of attack and airspeed is also presented. The Jetpack mathematical model
and the ducted-fan moment model have improved the physical understanding of the Martin Jetpack
which has lead to significant advances in the flight performance of the Jetpack.
2 Methodology
The first step in the development of the Jetpack model is to determine all the forces and moments
acting on the Jetpack. This is done by identifying and deriving suitable equations for all the key
features on the Jetpack, which include: two ducted-fans, Jetpack body, three sets of control vanes,
gyration moments from rotating parts and landing gear forces from ground contact. With the forces
and moments known or approximated, the Jetpack model can be developed following the conventional
2. Relate the forces and moments to accelerations via Newton’s second law.
3. Solve the system of ordinary differential equations to determine the system states.
Two coordinate systems need to be established, one for the Jetpack (body frame) located at the centre
of gravity and another for the inertial system (global frame). The reason for two coordinate systems
is that all the forces and moments must be solved with respect to the inertial system to account for
centripetal and gyration effects from the moving body frame. In the case of aircraft where most forces
are generated due to the geometry of the aircraft it is easier to account for forces with respect to the
body frame.
6
Euler angles, as described in [18, 19], are used to relate between the aircraft’s (body frame) co-
ordinate system (x, y, z) and the earth fixed (global frame) coordinated system (xg , yg , zg ), via two
intermediate coordinate systems (x1 , y1 , z1 ) and (x2 , y2 , z2 ). The directional cosine matrix, equation
(3), describes the transformation of a vector quantity from the earth fixed axis to the body fixed axis,
and is created by considering the positive right hand rotations of Ψ about the zg axis, then Θ about
cos Θ cos Ψ cos Θ sin Ψ − sin Θ
DCM =
sin Φ sin Θ cos Ψ − cos Φ sin Ψ sin Φ sin Θ sin Ψ + cos Φ cos Ψ sin Φ cos Θ
(3)
cos Φ sin Θ cos Ψ + sin Φ sin Ψ cos Φ sin Θ sin Ψ − sin Φ cos Ψ cos Φ cos Θ
x xg
y = DCM y (4)
g
z zg
To relate back from the body fixed axis to the earth fixed axis we simply reverse the calculation
by inverting the directional cosine matrix. As the directional cosine matrix is an orthogonal matrix
x
g x x
y = DCM−1 y = DCMT
g
y
(5)
zg z z
It is also necessary to relate angular quantities from the earth fixed system to the body fixed
system, this is done in a similar manner to the translational case, but transformation matrix is no
longer orthogonal, as the angular rotation rates are about their intermediate axes, as described in [19],
7
1 0 − sin Θ
DCMang =
0 cos Φ sin Φ cos Θ (6)
0 − sin Φ cos Φ cos Θ
p Φ̇
q = DCMang Θ̇ (7)
r Ψ̇
As the matrix is no longer orthogonal its inverse can be undefined, which has the implication that
π π
the transformation matrix can be either unsolvable or inaccurate. This occurs when Θ = 2 or Θ ≈ 2.
π
Hence whenever the simulated Jetpack has an attitude close to Θ ≈ 2 results will be inaccurate,
fortunately steady flight at this attitude is not physically achievable by the Martin Jetpack so the model
does not have to account for this. This could be overcome by using quaternion transformations[19].
Figure 3 shows the key forces acting on the Jetpack in flight in the longitudinal plane. Similar forces
are experienced in the lateral plane when subjected to motion within this plane.
Duct Thrust Figure 4 shows a cross-section of the air flow through a ducted-fan control volume
with the fan represented by an actuator disc (solid line). The actuator disc is used to model the
static pressure rise across the fan (p3 − p2 ) that in reality is created by a rotating propeller or fan.
Referring to the static pressure chart at the bottom of Figure 4, the presence of the fan decreases the
static pressure in-front of the fan and increases the pressure across the fan between stations 2 and 3
inducing a fluid flow. Behind the fan the static pressure is constrained by the duct to increase back to
atmospheric pressure, patm , which is the key difference between an open propeller and a ducted-fan.
The induced fluid velocity increases the initial velocity at station 1 to a higher final velocity at the duct
exit plane, station 4. In accordance to Newton’s law the change in the momentum between stations 1
8
x2,x
Θ Φ
x1 yg
xg Ψ Ψ y1,y2
Θ
Φ
q y
r Ψ
Φ
z Θ
z2
zg,z1
Figure 2: Shows the Euler linear and angular rate transformation from global frame to body frame.
The angular rate transformation has the rotational quantities p, q, r, Φ̇, Θ̇, and Ψ̇ superimposed on top
linear Euler transformation.
2Td
2Md
ld
2Fn
lD
D
lpitch
2Fpitch,L
W
2Fpitch,D
V0 α
Figure 3: Free body diagram showing forces acting on the Jetpack in the longitudinal/pitch plane.
9
Fan/rotor area, Ar Duct exit area, Ad
V4
Vr
V1
p1 p4 patm
Static Pressure
p3
p2
1 2 3 4
Following the methodology presented in [2], the derivation of the thrust model for a ducted-fan, as
where Vr is the absolute velocity at the rotor, Vri is the induced velocity at the rotor and ∆V is the
change in velocity between stations 1 and 4. Assuming air density is constant throughout the control
volume mass conservation implies that the change in velocity between stations 1 and 4 is:
Vr Vri + V1
∆V = V4 − V1 = − V1 = − V1 (9)
d d
Vr
Td = ṁd ∆V = ρAr Vr − V1 (10)
d
10
s
V 1 d V12 2d Td d
Vr = + + (11)
2 4 ρAr
s
1 Td 1 V1 3 Td2 V12 Td3
Pd = ṁd V42 − V12 = Vr ( + ) = Td V1 +
+ (13)
2 2 d Vr 4 16 4ρAr d
Equation (13) is rearranged to form a cubic function with Td as the variable. The cubic equation is
solved to find the smallest real root of Td , which becomes the ducted-fan thrust for the given duct
geometry, inflow velocity V1 , input power Pd and air density ρ. The power supplied to the ducted-fan
1
Pd = ηd Pe (14)
2
where the division of two is made as there are two ducted-fans, ηd is the ducted-fan efficiency, and
Pe is the power supplied by the engine. The ducted-fan efficiency is assumed to be constant, as no
appreciable loss in efficiency has been found from the test flights to date. The engine power is controlled
by engine throttle setting, which is automatically controlled through a control law, as per the Jetpack.
A first order transfer function is used, between throttle setting and engine power, to model engine
latency, due to engine inertia, where a realistic time constant was found by trail.
Ducted-Fan Normal Force For flight regimes other than hover, vertical climb and descent there
will be a velocity component perpendicular to duct. This perpendicular velocity component is absorbed
by the duct as the ducted-fan has the effect of straightening any perpendicular airflow, as shown in
Figure 5. Hence, the perpendicular component of the free-stream air momentum is redirected along
the axis of the duct. This results in a force being created, ducted-fan normal force, due to the change
in momentum of the air. The orientation of the ducted-fan normal force Fn is shown on Figure 5,
11
where it can be seen that this force has a large drag component for VTOL aircraft applications such
as the Martin Jetpack. The ducted-fan normal force is calculated in [20] as:
where ṁd is the mass flow through the ducted-fan, calculated from equation (11). V0 is the free-stream
velocity and α is the angle of attack defined as the angle between the airspeed vector and the duct
where V0,x and V0,y are the airspeed in the x and y directions relative to the Jetpack, respectively.
T
Fd = [Fn,x , Fn,y , −Td ] (18)
where Td is the duct thrust component acting along the duct axis. The net duct force Fd is acting at
the duct quarter chord point as shown in Figure 5. The duct quarter chord point has been chosen as
the reference point as this is the historic reference point for ducted-fans.
Duct Centre of Pressure Model A method has been developed to model the duct centre of
pressure based on the physical phenomenon of the air mass turning through the duct. If the duct is
at some angle of attack to the free stream velocity, due to the presence of the ducted-fan then the air
mass passing through the duct control volume must turn through an angle, which is assumed equal to
the angle of attack, refer to Figure 6. Two vectors V0 and V4 can be drawn at the centre of the entry
and exit of the control volume, as shown Figure 6. The centre of the arc connecting these two vectors
is essentially the centre of pressure, as this point represents the turning point for the air mass.
If the exit vector is aligned to the duct axis, which can be assumed if the flow is uniform through
12
α Td
Td
V0
Fn
V0,z = V1 V0,x
Fn
= Md
xg
zg
α
V0
Fn
r r CPd
α α/2
V4
13
the duct, the entry vector requires constraints to fully define the location of the duct centre of pressure.
The entry vector position can be defined based on the known mass flow rate and free stream velocity
– using the conservation of mass – the radius r of the stream tube at the entry of the control volume
r
ṁ
r= (19)
πρV0
where ṁ is the mass flow rate through the duct and V0 is the free stream velocity. The duct centre
of pressure CPd located along the duct axis and measured from the duct quarter chord, as shown on
where α is the angle of attack the duct. This is done for both x and y axes so that the duct centre of
It can be seen be inspection of Figure 6 and equations (19) and (20) that the height of the duct
centre of pressure above the duct chord line increases as airspeed diminishes towards zero and angle
The duct moment about the intersection of the duct axis and duct quarter chord line can be
calculated by the cross product of the duct centre of pressure vector, CPd , and the net duct force
vector, Fd , as:
Md = CPd × Fd (22)
Md,CG = ld × Fd + Md (23)
As there are two ducts, this is done for both left and right hand ducts so that:
14
Md,RH,CG = ld,RH × Fd,RH + Md,RH (24)
The control about the Jetpack’s longitudinal x, lateral y, and vertical z axes is achieved by deflection
of the roll, pitch, and yaw vanes respectively. Each axis has a pair of control vanes that hinge about the
vane centre of pressure (quarter chord). All vane lift and drag forces were experimentally determined
for a range of vane angles from zero to stall angle of attack. This was accomplished by measuring the
vane lift and drag reactions in the slipstream of a single ducted-fan test apparatus that uses the same
ducted-fan as used the Jetpack. Using load cells to measure the vane lift, drag, and duct thrust the
lift and drag coefficients for the control vanes were determined by the following relations:
Lvane
CL,vane = 1 2
(26)
2 ρVd Avane
Dvane
CD,vane = 1 2
(27)
2 ρVd Avane
where: Vd = V4 is the duct exit velocity and is calculated using equations (8) and (11) for the
measured duct thrust, Avane is the vane planform area, and Lvane and Dvane are the measured lift
and drag force at each angle of attack. Vane lift and drag coefficients are used to model the vane
reactions as a function the duct velocity which is a function of the duct thrust. Using this method
the simulated control forces and moments become a function of the duct thrust, which replicates the
Jetpack prototype. The experimentally measured lift and drag coefficients versus vane angle of attack
are shown in Figure 7, where the major and minor axis of the crucifix indicate the measurement
uncertainty. The stall angle was found to be at approximately 30◦ of deflection, which is typically for
15
1.2 Lift
Drag
0.6
0.4
0.2
0
0 5 10 15 20 25 30
Angle of Vane Deflection [°]
Figure 7: Lift and drag coefficients versus angle of attack for control vanes
Roll Vanes Two roll vanes are used to generate roll moments to orientate the aircraft about the x
axis. When deflected from their neutral position (ξ = 0), the roll vanes generate an aerodynamic lift
1 1
Froll,RH = ξf [0, ρVd2 Aroll CL,vane (ξ) , ρVd2 Aroll CD,vane (ξ)] = Froll,LH (28)
2 2
where ξf is the roll vane modifying factor used to adjust the dynamic pressure due to the position of
the control vane within the ducted-fan wake, as the dynamic pressure experienced by the control vanes
is not uniform between all the vanes. Due to the perpendicular distance between roll vane centre of
similarly the left hand roll vane produces the following moment:
where the position vector lroll,LH describes the distance from the CG to the centre of pressure of the
16
Pitch Vanes Similarly to the roll vanes the pitch vanes also create moments about their intended
axis, the lateral y axis. The pitch vanes create lift forces in the x direction and drag forces in the z
direction.
1 1
Fpitch,RH = ηf [ ρVd2 Apitch CL,vane (η) , 0, ρVd2 Apitch CD,vane (ξ)] = Fpitch,LH (31)
2 2
Yaw Vanes The yaw vanes on the Jetpack act differential to one another, similar to ailerons on an
airplane, to produce a moment about the z axis. Due to the differential nature of the yaw vanes a
1 1
Fyaw,RH = ζf [ ρVd2 Ayaw CL,vane (ζ) , 0, ρVd2 Ayaw CD,vane (ζ)] = −Fyaw,LH (34)
2 2
Net Control Forces The net control forces can now be calculated as the sum of the individual vane
forces:
17
similarly for net control moments:
The aerodynamic forces on the Jetpack like most aircraft are a function of the airspeed. The airspeed
is calculated as the difference between ground speed of the aircraft and the wind speed as:
V0 = V − Vwind (39)
The Jetpack can be considered to be a bluff body, as it is not streamlined in anyway, and for this
reason it is assumed that the body itself does not create any lift forces. Also, as the Jetpack’s intended
velocity range is from 0 to 100 km/h any lift force from the body would be small due to the low
dynamic pressure. Likewise aerodynamic moments from the Jetpack body are also neglected. Hence,
drag forces are significant and are accounted for by the following equation:
C A kV k
D,x x 0,x 0,x V
1
FD = − ρ CD,y Ay kV0,y k V0,y (40)
2
CD,z Az kV0,z k V0,z
where CD is the drag coefficient, A is the projected area and V0 is the airspeed. The subscripts x, y, z
denote the axis the fore mentioned variables refer to. The moment due to the distance between the
MD = lD × FD (41)
Takeoff and landing are critical flight phases that the Jetpack pilot must be proficient at. Hence,
a ground interaction model was added to increase the realism of the Jetpack for pilot-in-the-loop
simulations. This was done by modeling each ground contact point as a spring and damper system
18
x
z
Body Frame
MLG
xg
zg
Global Frame
Ground
hLG
FLGParallel,g,x
FLGNormal,g,z
with the spring force normal to the horizontal ground and damping forces normal and parallel to the
ground. The landing gear forces were modeled as a compression only spring, so that the landing gear
force only occurs when the contact point of the landing gear is below ground height, as shown on Figure
8. When a contact point is below ground height landing gear reaction forces normal and parallel to
the surface are calculated. The orientation of these forces are defined in the global frame, as shown on
Figure 8. The ground reaction or normal force to the surface, FLG,z,g , is proportional to the distance
hLG . The drag forces, FLG,x,g and FLG,y,g , which prevent the Jetpack from sliding along the ground
are parallel to the surface and in the opposite direction and proportional to the motion of the Jetpack.
The Jetpack landing gear consists of three contact points, but a forth contact point was also added.
This extra contact point located at the top of the Jetpack is used to prevent the Jetpack from inverting
in the simulation environment. Hence, the forth point gives the Jetpack a simple three dimensional
tetrahedral shape in the simulation environment. For each contact point the following was determined:
where: hLG is the height of the landing gear contact point above ground, zg is the height of the Jetpack
above earth fixed axis, lLG,g,z is the z component of the distance vector from the centre of gravity to
19
landing gear contact point, and ground is the height of ground above global frame.
where: uLG,g is the velocity vector of the contact point with respect to the earth fixed axis, ug is
the velocity of the Jetpack with respect to earth fixed axis, DCM−1 is the inverse Directional Cosine
Matrix, ω is the angular body rates and lLG,g is the distance vector from the CG to the contact point.
where K is a spring constant, Cz is the damping coefficient normal to the ground, and uLG,g,z is the
z component of the landing gear velocity in the earth fixed axis. Values of K and C were chosen to
• The landing gear force parallel to the surface and opposite to the direction of motion (sliding
drag):
−Cx uLG,g,x
hLG > 0
FLG,x,g = (45)
0
hLG ≤ 0
where: Cx is the damping coefficient parallel to the ground in the x direction, uLG,g,x is the x velocity
component of the landing gear in the global frame. The force FLGP,y,g is calculated using the same
The landing gear force for each contact point in terms of the Jetpack body frame is:
Moments are calculated for each individual landing gear contact point by multiplying the distance
vector from the CG to the contact point with its force as:
20
MLG = lLG × FLG (47)
The total combined landing gear forces and moments is the sum of the individual landing gear
The Jetpack has significant rotating inertia (angular momentum) due to the rotating components at
high rotation speed, these include: two fans rotors, drive belts and rotating engine parts. The gyration
moment is proportional the pitch and roll rate on the Jetpack and induces a motion similar to that of
a spin top. For example, using the right hand rule, the Jetpack’s rotating parts all rotate around the
−z axis and if an angular downwards pitching rate, −q about y axis, is present a gyroscopic reaction
moment is induced creating positive roll about the x axis. This motion about the x axis then induces
a positive moment about the y axis which in turn produces a negative moment about the x axis.
The cyclic motion continues until it is dampened. To simply summarize: downwards pitch leads to
right roll; right roll leads to upwards pitch; upwards pitch to left roll; left roll to downwards pitch to
Where: ω is the Jetpack rate vector [p, q, r]T ; Irot is the inertia of the rotating parts, as determined
from computer aided design (CAD) geometry, and Ωrot is the angular velocity of the rotating parts.
2.3 Kinematics
With the key forces and moments on the Jetpack described mathematically the summation of these is:
21
Employing Newton’s Second law F = ma for a rigid body with six degrees of freedom (x, y, z, φ, θ, ψ)
and taking into account the kinematics for the translational motion in x, y, z we get:
F
a= − ω × u + DCMg (51)
m
where: m is the mass of the aircraft, ω is the vector of angular rates p, q, r, DCM is the Euler
transformation matrix and g is the acceleration due to gravity resolved in the aircraft frame.
Similarly, using Newton’s Second law for the rotational degrees of freedom Φ, Θ, Ψ we get:
where I is the inertia tensor of Jetpack determined from CAD geometry. The ODEs described in
equations (51) and (52) are solved using Matlab Simulink software, which is optimised to solve dynamic
state-space problems.
2.4 Controllers
Due to the unstable or at best neutrally stable [21] response of VTOL aircraft and particularly the
Jetpack with its inherently low inertia– a fly-by-wire is necessary to introduce artificial stability to
achieve controlled flight. For the same reason the Jetpack model also requires artificial stabilization
via feedback loops and controllers. Beneficially, the model allows for various types of controllers to be
economically evaluated. The simplest controller found to achieve desirable handling qualities was the
rate controller. However, to aid in remote control test flying of the Jetpack a nested attitude controller
The rate controller, Figure 9, corrects the aircraft’s angular rates p, q, r as well as the climb rate wg
22
cmd(t) ε(t) U(t) x(t)
K system
-
where: (t) is the error, cmd(t) is the commanded signal from the pilot’s joystick, and x(t) is the
corresponding rate value representing p, q, r and wg . The system input signal U (t) is calculated in
terms of vane deflections ζ, η, ξ or thrust setting Fthrust by multiplying the error with an appropriate
Tuned gain values, K, were quickly determined by pilot in the loop simulations.
The nested or cascaded controller contains two loops, an outer and an inner. The inner loop is used
to control faster or higher order dynamics while the outer loop controls the overall system, which is
usually a slower dynamic. In the case of the Jetpack model the outer loop controls roll Φ and pitch Θ
attitudes, while the inner loop controls the angular roll p and pitch q rates. The error from the outer
loop (difference between attitude target and actual attitude, equation (55)) is processed by the outer
loop controller, equation (56), and produces a rate target, Uo (t). A saturation (limiter), equation (57),
is used to prevent too large a rate target from being passed to the inner loop. The difference between
rate target and actual rate, equation (58), is then processed by the inner controller, equation (59), to
produce the system input Ui (t), which is in terms of roll and pitch vane deflection.
23
cmdo(t) εo(t) Uo(t) cmdi(t) εi(t) Ui(t)
Ko saturation Ki system
- -
x(t)
x(t)
U
o kUo k < Limitu,o
cmdi (t) = (57)
LimitU,o Uo
kUo k kUo k > Limitu,o
3 Model Validation
The model was validated in two ways: comparison of steady state data and comparison of dynamic data.
Both methods compared simulation data directly to flight data collected from flight tests performed
by the Jetpack. Initial analysis of the model indicated that the original Jetpack concept, Figure 1,
needed to be modified to that shown in Figure 11 to improve the flight performance and handling of
the Jetpack, as controlled flights in outdoor environments of the original concept was not possible. The
modifications mainly involved lowering the ducted-fans in relation to the aircraft’s centre of gravity
to better align the the duct centre of pressure to center of gravity. This significantly improved the
aircraft’s performance and allowed for flight tests to be preformed over a range of airspeeds from 0 to
∼ 60 km/h. All flight tests were performed: unmanned by radio control, during calm conditions (wind
24
Engine
Yaw vane
Roll vane
Rear
landing leg
Pogo
Pitch vane
The steady state comparison is made by solving for the trimmed (equilibrium) level-flight pitch vane
deflection and airspeed for a given pitch attitude, and comparing this result with time averaged flight
data from Jetpack flights in the longitudinal trim condition. Figure 12 shows the comparison between
simulated and collected flight data. It can be seen, on Figure 12, that the simulated airspeed is higher
than that of the actual Jetpack for a given pitch attitude indicating that the drag coefficient terms in
the model should be increased. However, it must also be noted that the recorded airspeed from the
flight data at higher pitch down angles has not reached steady state, as the airspeed is still accelerating.
High steady state airspeeds have not yet been obtained due to practicality considerations of flying the
The predicted pitch vane deflection shows good relation with that measured from the test flight.
This indicates that the duct centre of pressure and moment model described herein is a valid method
to predict the duct, and hence, the Jetpack’s behaviour up to a pitch down attitude of 30◦ . Further
25
flight tests expanding at higher pitch down attitudes will need to be performed to validate the model
for greater attitudes. Both the prototype flight data and model show that the pitch vane deflection
initial rises and reaches a maximum at approximately 10◦ pitch down attitude and decreases passing
through neutral vane position at approximately 27◦ pitch down attitude. This rise in vane deflection
corresponds to a negative pitch vane moment (as per sign convention, Figure 1) being created by the
pitch vanes to oppose the positive pitch moment of the Jetpack, which is largely due to duct reactions.
The point where the trim vane deflection is zero (neutral position) indicates that the aerodynamic
centre of pressure is aligned with the centre of gravity at this pitch attitude and airspeed. Ideally, the
pitch vanes should be at their neutral position throughout the flight range to maximize the control
authority for manoeuvering. However, as is evident from the flight test results the centre of pressure
changes as a function of attitude and airspeed, hence the control vanes must also deflect to create
25
Airspeed Model
Pitch Vane Model
Airspeed Prototype
20 Pitch Vane Prototype
Airspeed [m/s] & Pitch Vane Deflection [°]
15
10
−5
0 5 10 15 20 25 30
Pitch Down Attitude [°]
Figure 12: Comparison between model and prototype for trimmed level-flight in the longitudinal plane.
26
3.2 Dynamic Comparison
A dynamic comparison was made to validate that the model shows a realistic dynamic response. This
was done by using the pitch command recorded from a flight test as the input pitch command for the
model. A 60 s test flight beginning in a steady hover with steadily increasing and abrupt pitch attitude
commands was performed to record the pitch commands and corresponding pitch attitude response of
the Jetpack. The comparison of the pitch attitude between the model and the prototype is shown on
Figure 13. It can been seen that the model’s response follows the trend of the pitch command, but
typically undershoots the trend measured from the prototype. The differences between simulation and
• Overestimation of the Jetpack’s inertia, which reduces the response of the model.
0
Pitch Attitude [deg]
−5
−10
−15
−20
0 10 20 30 40 50 60 70
Time [s]
Figure 13: Comparison of dynamic response in pitch attitude for given pitch command
27
4 Conclusion and Final Comments
A mathematical model has been developed and successfully used to model the behaviour of a twin
ducted-fan VTOL aircraft, namely the Martin Jetpack. Following the conventional aircraft modeling
methodology, mathematical formulas were developed to model the physics of the key Jetpack features.
This includes modeling the forces and moments produced by both ducted-fans, the control vanes,
the body aerodynamics, and the gyration moment. Landing gear forces are also included to allow
the Jetpack model to simulate the ground interaction during takeoff and landing, which is a vital
phenomenon to simulate. Combining all of these forces and moments and employing Newton’s second
law of motion, Euler angle transformations, and the kinematic equations, for a six degree of freedom
system, a system of ODEs was produced. The ODEs are solved, in real-time, using Matlab-Simulink
Similarly to the actual Jetpack the model also requires artificial control to stabilize and obtain
humanly controllable flight. It was found that the rate controller was the simplest controller to give
desirable handling qualities. A nested attitude-rate controller was also added to the model to reflect
The model presented includes a novel analytical method to predict the duct centre of pressure as a
function of airspeed and angle of attack, which allows for the duct pitching moment to be accurately
predicted.
Future work will focus on improving the accuracy of the model by means of parameter identification
and/or direct measurement of the duct aerodynamic forces and moments. The model presented can
now be used for fundamental aerodynamic design, control system design and pilot in the loop flight
simulation.
Acknowledgments
The authors gratefully acknowledge Martin Aircraft Company for their technical support during flight
experimentation.
28
Funding Acknowledgments
This work was supported by the Foundation for Research, Science and Technology New Zealand [grant
number MAIR0802].
Nomenclature
a Acceleration [m/s2 ]
D Drag [N]
F Force [N]
29
Fpitch Pitch vane control force [Fpitch,L , 0, Fpitch,D ] N
L Lift [N]
M Moment [Nm]
m Mass 220 kg
30
P Engine Power 200 HP
p Pressure
U Input signal
u, v, w Velocity in x, y, z directions
V Velocity [m/s]
W Weight [N]
x, y, z x, y, z axis respectively
Subscripts
d Ducted-fan
o Outer
31
p Open propeller
r Rotor
Note all forces, moments and vectors refer to body fixed axis unless otherwise stated.
Greek Letters
α Angle of attack
ηd Duct efficiency
CG Center of gravity
CP Centre of pressure
cmd Command
32
V T OL Vertical takeoff and landing
LG Landing gear
T OM Takeoff mass
References
[1] Jr. Barnes W, McCormick. Aerodynamics of V/STOL Flight. Academic Press INC., dover edition
edition, 1967.
[2] Jason L. Pereira. Hover and Wind-Tunnel Testing of Shrouded Rotors for Improved Micro Air
[3] Jonathan Fleming and Troy Jones. Improved Control of Ducted Fan VTOL UAVs in Crosswind
[4] Will E. Graf, Jonathan Fleming, and Wing Ng. Improving Ducted Fan UAV Aerodynamics in
[5] R.J. Weir. Aerodynamic design considerations for a free flying ducted propeller. In AIAA Atmo-
[6] R. Allan Wallis. Axial Flow Fans and Ducts. A Wiley-Interscience publication, 1983.
[7] Preston Martin and Chee Tung. Performance and Flowfield Measurements on a 10-inch Ducted
Rotor VTOL UAV. In 60th Annual Forum of the American Helicopter Society, June 2004.
[8] Paul F. Yaggy and Kenneth W. Goodson. Aerodynamics of a Tilting Ducted Fan Configuration.
33
[9] Paul F. Yaggy and Kenneth W. Mort. A Wind Tunnel Investigation of a 4 Foot Diameter Ducted
Fan Mounted on the Tip of a Semispan Wing. Technical Report D-776, NASA, March 1961.
[10] Kenneth W. Mort and Paul F. Yaggy. Aerodynamic Characteristics of a 4-Foot-Diamter Ducted
Fan Mounted on the Tip of a Semispan Wing. Technical Report D-1301, NASA, Ames Research
[11] Robert J. Tapscott and Henery L. Kelley. A Flight Study of the Conversion Manoeuver of a
[12] Arthur C. Robertson, Joseph Stuart III, Palo Alto, and Robert A. Wagner. Vertical Take-Off
[13] Jonathan Fleming, Troy Jones, and Wing Ng. Improving Control System Effectiveness for Ducted
Fan VTOL UAVs Operating in Crosswinds. In 2nD AIAA "Unmanned Unlimited" Systems, Tech-
[14] Sighard F. Hoerner. Fluid-Dynamic Lift. Hoerner Fluid Dynamics, 2n ed. edition, 1985.
[15] Osgar John Ohanian III. Ducted Fan Aerodynamics and Modeling, with Applications of Steady
and Synthetic Jet Flow Control. PhD thesis, Virginia Polytechnic Institute and State University,
2011.
[16] Mark De Roche. Thrust augmentation and control of a ducted-fan vtol air vehicle. In AHS Future
[17] Eric N. Johnson and Michael A. Turbe. Modeling, Control, and Flight Testing of a Small Ducted
Fan Aircraft. Journal of Guidance, Control, and Dynamics, 29, July-August 2006.
[19] Bandu N. Pamadi. Performance, Stability, Dynamics, and Control of Airplanes, Second Edition.
34
[20] Andy Ko, Osgar John Ohanian, and Paul Gelhausen. Ducted Fan UAV Modeling and Simulation
in Preliminary Design. In AIAA Modeling and Simulation Technologies Conference and Exhibit,
2007.
[21] Paul J, Weitz. A Qualitative Discussion of the Stability and Control of VTOL Aircraft During
Hover (Out of Ground Effect) and Transition. Master’s thesis, United States Naval Postgraduate
List of Figures
2 Shows the Euler linear and angular rate transformation from global frame to body
frame. The angular rate transformation has the rotational quantities p, q, r, Φ̇, Θ̇, and Ψ̇
3 Free body diagram showing forces acting on the Jetpack in the longitudinal/pitch plane. 9
7 Lift and drag coefficients versus angle of attack for control vanes . . . . . . . . . . . . . 16
12 Comparison between model and prototype for trimmed level-flight in the longitudinal
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
35
Radiator ducting
Ducted-fan
Engine
v,y
M,q
Pilot
Pitch/roll/yaw
joystick Yaw vane
ζ
Pilot display
Throttle
joystick
Pogo, main
N,r landing gear
w,z
36
x2,x
Θ Φ
x1 yg
xg Ψ Ψ y1,y2
Θ
Φ
q y
r Ψ
Φ
z Θ
z2
zg,z1
37
2Td
2Md
ld
2Fn
lD
D
lpitch
2Fpitch,L
W
2Fpitch,D
V0 α
38
Fan/rotor area, Ar Duct exit area, Ad
V4
Vr
V1
p1 p4 patm
Static Pressure
p3
p2
1 2 3 4
39
α Td
Td
V0
Fn
V0,z = V1 V0,x
Fn
= Md
xg
zg
40
α
V0
Fn
r r CPd
α α/2
V4
41
1.2 Lift
Drag
1
Lift and Drag Coefficient
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30
Angle of Vane Deflection [°]
42
x
z
Body Frame
MLG
xg
zg
Global Frame
Ground
hLG
FLGParallel,g,x
FLGNormal,g,z
43
cmd(t) ε(t) U(t) x(t)
K system
-
44
Engine
Yaw vane
Roll vane
Rear
landing leg
Pogo
Pitch vane
45
25
Airspeed Model
Pitch Vane Model
Airspeed Prototype
20 Pitch Vane Prototype
Airspeed [m/s] & Pitch Vane Deflection [°]
15
10
−5
0 5 10 15 20 25 30
Pitch Down Attitude [°]
46
Pitch Att Comparison of Sim to FD: 2011.09.09.Log.09.49 660 to 720s
10
Simulated
Flight Data
Pitch Command
5
0
Pitch Attitude [deg]
−5
−10
−15
−20
0 10 20 30 40 50 60 70
Time [s]
47