2009 American Control Conference ThA13.

4
Hyatt Regency Riverfront, St. Louis, MO, USA
June 10-12, 2009

Automatic Path Planning and Control Design for Autonomous

Landing of UAVs using Dynamic Inversion
Shashiprakash Singh and Radhakant Padhi

Abstract— In this paper a nonlinear control has been de- and unmanned aircrafts. Linearized model of the aircraft have
signed using the dynamic inversion approach for automatic been used in the literature for autolanding using separate
landing of unmanned aerial vehicles (UAVs), along with associ- longitudinal and lateral dynamics, where the control is de-
ated path planning. This is a difficult problem because of light
weight of UAVs and strong coupling between longitudinal and signed using linear matrix inequality method [2]. Modern
lateral modes. The landing maneuver of the UAV is divided control methods like H2 /H∞ have also been used for landing
into approach, glideslope and flare. In the approach UAV of UAVs [3]. However, linear system based approaches have
aligns with the centerline of the runway by heading angle a strong limitation that they work within a small operating
correction. In glideslope and flare the UAV follows straight range. Gain scheduling can perhaps be used to overcome this
line and exponential curves respectively in the pitch plane
with no lateral deviations. The glideslope and flare path are limitation to a limited extent. However gain scheduling is a
scheduled as a function of approach distance from runway. tedious process and there is no guarantee that the interpolated
The trajectory parameters are calculated such that the sink gains can assure stability of the closed loop system [4].
rate at touchdown remains within specified bounds. It is also In this paper a nonlinear control has been designed using
ensured that the transition from the glideslope to flare path the dynamic inversion technique [5] for automatic landing of
is smooth by ensuring C1 continuity at the transition. In the
outer loop, the roll rate command is generated by assuring a co- UAVs. Dynamic inversion relies on the philosophy of feed-
ordinated turn in the alignment segment and by assuring zero back linearization. This feedback control structure cancels
bank angle in the glideslope and flare segments. The pitch rate the nonlinearities in the plant such that the closed loop plant
command is generated from the error in altitude to control the behaves like a stable linear system. This method has several
deviations from the landing trajectory. The yaw rate command advantages, like simplicity in the control structure, ease of
is generated from the required heading correction. In the inner
loop, the aileron, elevator and rudder deflections are computed implementation, global exponential stability of the tracking
together to track the required body rate commands. Moreover, error etc. Note that a feedback linearization technique has
it is also ensured that the forward velocity of the UAV at the been successfully demonstrated through simulation for auto-
touch down remains close to a desired value by manipulating matic landing of a high performance aircraft [1]. The control
the thrust of the vehicle. A nonlinear six-DOF model, which implemented for the UAV in this paper has inner loop and
has been developed from extensive wind-tunnel testing, is used
both for control design as well as to validate it. outer loop structure. The outer loop converts the guidance
commands to body rates. Where as inner loop generates the
I. INTRODUCTION control to track the body rates desired by outer loop. There
The capabilities of UAV as flying machines can be ex- is a separate loop for velocity control which regulates the
ploited to carry out surveillance missions and remote oper- desired velocity by throttle control.
ations. The recovery of UAVs in landing is one of the key The landing maneuver of the UAV is divided into ap-
operations in flight which define the overall success of the proach, glideslope and flare. The requirement of variable
mission. Landing becomes a more challenging task in case gain and blending function [6] for gains at the transition
of light weight vehicles such as UAVs, mainly because of of glideslope and flare have been reported in literature.
of the fact that the longitudinal and lateral modes in UAVs This requirement has been overcome by careful selection of
are coupled. This problem cannot be dealt effectively with landing trajectory parameters. This ensures that the transition
linearized models, where underlying assumption is that the from glideslope to flare path is smooth. The glideslope and
longitudinal and lateral modes are decoupled. In this paper a flare path parameters are computed online (i.e. they are
nonlinear control design approach is proposed using dynamic not fixed apriori). Note that the trajectory parameters are
inversion for automatic landing of UAVs. calculated such that the sink rate at touchdown remains
Landing trajectories of aerial vehicles typically consists within specified bounds.
of approach, glideslope and flare [1]. A successful landing The proposed control design has been experimented with
would depend upon the good selection of landing trajectory respect to a UAV built in the UAV lab of Indian Institute
and closely following it. Various methods have been adapted of Science. Complete nonlinear six degree of freedom (six-
to tackle the problem of automatic landing both for manned DOF) equations of motion are used to define the dynamics
of the UAV. The aerodynamic coefficients for force and
Shashiprakash Singh, Graduate Student, is with Department of Aerospace moments are found from the curve fitting of the wind tunnel
Engineering, Indian Institute of Science, Bangalore 560012, India. data [7], the philosophy of which is available in the literature
Radhakant Padhi (Contact author), is Asst. Professor at Department of
Aerospace Engineering, Indian Institute of Science, Bangalore 560012, India [8]. Note that while curve fitting the nonlinearities in the
[email protected] aerodynamic coefficients are not neglected.

978-1-4244-4524-0/09/$25.00 ©2009 AACC 2409

 II. M ATHEMATICAL M ODEL Aerodynamic coefficients obtained from curve fitting on
wind tunnel data [7] are given as
Simulation requires an accurate mathematical model of
the dynamical system so that analysis be meaningful. Here CX =CX0 +CXα (α )α +CXδ e (α )δ e +CXQ (α )Q̄
nonlinear six-DOF equations are used for the simulation. The CY =CYβ (α )β +CYδ a (α )δ a +CYδ r (α )δ r +CYP (α )P̄ +CYR (α )R̄
aerodynamic force and moment coefficients are found from CZ =CZ0 +CZα (α )α +CZβ β +CZδ e δ e +CZQ (α )Q̄
curve fitting on wind tunnel data of the UAV.
Cl =Clβ (α )β +Clδ a (α )δ a +ClP (α )P̄ +ClR (α )R̄
Cm =Cm0 +Cmα (α )α +Cmβ (α , β )β +Cmδ e (α )δ e +CmQ (α )Q̄
A. Six-DOF equations of motion
Cn =Cnβ (α )β +Cnδ r (α )δ r +CnP (α )P̄ +CnR (α )R̄
Under the assumptions of airplane to be a rigid body and
earth to be flat the complete set of six-DOF equations [9] of where,
1
motion are given by following differential equations. [P̄ Q̄ R̄] = [bP cQ bR]
2Vt
U̇ = RV − QW − gsinθ + Xa + Xt (1) In above equations α is angle of attack and β is sideslip
V̇ = PW − RU + gsinφ cosθ +Ya (2) angle. δ a, δ e, δ r are aileron elevator and rudder control
deflections respectively. Some of the static and dynamic
Ẇ = QU − PV + gcosφ cosθ + Za (3) derivatives are functions of α and β . The detailed functions
Ṗ = c1 RQ + c2 PQ + c3 La + c4 Na (4) and constants are given in appendix. Only static tests were
Q̇ = c5 PR + c6 (P2 − R2 ) + c7 (Ma + Mt ) (5) conducted in wind tunnel. The dynamic derivatives are found
Ṙ = c8 PQ − c2 RQ + c4 La + c9 Na (6) with Athennae Vortex Lattice (AVL) software [10].
φ̇ = P + Qsinφ tanθ + Rcosφ tanθ (7) C. The UAV
θ̇ = Qcosφ − Rsinφ (8) Numerical simulations are carried out with the data of a
ψ̇ = Qsinφ secθ + Rcosφ secθ (9) prototype UAV, named as all electric airplane-2 (AE-2) [7]. It
is designed and developed at UAV lab of aerospace engineer-
ḣ = Usinθ −V sinφ cosθ −W cosφ cosθ (10) ing department, IISc. AE-2 (Fig. 1) is a fixed wing airplane
designed for autonomous flying. It has been designed for
In above equations U,V,W are velocity components and long endurance, where the flight duration is expected to be
P, Q, R are roll, pitch and yaw rates respectively about the around 45 minutes. The thrust generating unit of the AE-
body axis. φ , θ , ψ are euler angles and h is the height above 2 is an electric motor with propeller, which is powered by
ground. x is forward distance and y is sideward distance. lithium-polymer battery. It has a pusher configuration of
x and y position equations are not written above, but they thrust, this has been done so that gimballed camera can be
are included in the simulation. Xa ,Ya ,Za are the aerodynamic mounted at the nose. The physical data of AE-2 is given in
forces per unit mass and La ,Ma ,Na are the aerodynamic Table I.
moments about the body axis. Xt is the force per unit mass
in direction X due to thrust and Mt is the moment around TABLE I
the Y axis caused by thrust offset from the center of gravity PHYSICAL DATA OF AE-2
b c m Ixx Iyy Izz Ixz
(CG) of the aircraft. m m kg kgm2 kgm2 kgm2 kgm2
2 0.3 6 0.5062 0.89 0.91 0.0015
B. Aerodynamic forces and moments
The aerodynamic forces and moments are given by

qS
[Xa Ya Za ] = [CX CY CZ ] (11)
m
[La Ma Na ] = qS[bCl cCm bCn ] (12)
1
Xt = (Tmax σt ) (13)
m
Mt = −d (Tmax σt ) (14)

where q is dynamic pressure and S is wing planform area.

b is wing span and c is chord length. Tmax is the maximum
thrust (15 N) which can be produced by the electric motor
and propeller assembly. σt is throttle control which varies
from 0 to 1. It is assumed that thrust produced has linear
relation with throttle input. d is offset of the thrust line from Fig. 1. AE-2 (All Electric airplane-2)
the CG of the vehicle.

2410
 III. PATH P LANNING AND G UIDANCE B. Glideslope
The autonomous landing of UAVs requires a good path During glideslope the height of the vehicle above ground
planning and guidance. Here the landing is divided into three is controlled. The desired height is scheduled as a function
phases, which are approach, glideslope and flare. The desired of forward distance. The glideslope is a straight line path
trajectory is made a function of forward distance. whose slope is defined by the flight path angle.
A. Approach
During approach the vehicle should come and align with
the runway at a specified height from where the glideslope (x , h )
g g h
can be started.

N
y W
ψ0 E
S (x , h )
f f
(x ,y )
0 0
∗ (x , h )
x γ td td x
runway
(0,0) (xg0, hg0) (0, 0)
(x , y ) (x ,h )
g g ∞ c

Fig. 3. Landing geometry in x-h plane

In Fig. 3 (xg , hg ) is the point where glideslope begins.

(x f , h f ) is point where flare begins. (xtd , htd = 0) is the
Fig. 2. Approach geometry in x-y plane touchdown point. (0, 0) is the origin of inertial frame at the
beginning of runway. (xg0 , hg0 = 0) is a fictitious point on
In Fig. 2 (x0 , y0 ) is the initial point where UAV is at the ground from where glideslope is projected. (x∞ , hc ) is the
beginning of landing command. (xg , yg = 0) is point from final point of flare path, this point is chosen to be below
where glideslope begins. ψ0 is the initial orientation of the ground so that exponential flare path intersects the ground at
vehicle. The equation for path of approach can be written as touchdown point.
a straight line function of forward distance. When the UAV enters in glideslope initiation zone the

yg − y0
 required flight path angle which vehicle has to follow is
∗
y = y0 + (x − x0 ) (15) calculated as follows
xg − x0 
hg − hg0

We can find the desired heading angle to be followed by γ ∗ = tan−1 (21)
xg − xg0
inverting the kinematics. For this we first write a first order
error differential equation for y, where error ey = y − y∗ Than the desired height at every point of glideslope can be
written as a function of forward distance
e˙y + ky ey = 0 (16)
h∗ = (x − xg0 ) tanγ ∗ (22)
The solution of the above differential equation is given as
ey = ey0 e−ky t . If ky is chosen positive than ey → 0 as t → ∞. C. Flare
Expanding (16) we can write,
During flare the height of the vehicle above ground is
(ẏ − y˙∗ ) + ky (y − y∗ ) = 0 (17) controlled. The flare path is an exponential curve. The
desired height during exponential flare can be scheduled as
ẏ = y˙∗ − ky (y − y∗ ) (18) a function of forward distance.
From six-DOF equations we can write,
h∗ = hc + (h f − hc )e−kx (x−x f ) (23)
ẏ = ay sinψ + by cosψ (19)
The unknowns in above equation are flare height h f , distance
where, at which to begin flare x f , final height below the ground
ay , Ucosθ +V sinφ sinθ +W cosφ sinθ where flare trajectory should end hc , and constant kx . We
can solve for these four unknowns under the following four
by , V cosφ −W sinφ
constraints,
Substituting (19) in (18) than solving for ψ Initial condition: The point where glideslope ends and flare
begins be coincident. Substituting x = x f in (22) and (23)
 
˙∗ − k (y − y∗ )
 
y y  − tan−1 by than equating
ψ ∗ = sin−1  q (20)
a2y + b2y ay
hf = −(x f − xg0 ) tanγ ∗ (24)

2411
Initial slope: The slope at the beginning of flare and at the 2) Altitude control: The height during landing is con-
end of glideslope be same. Differentiating (22) and (23) than trolled by generating pitch command from the error in height.
equating at x = x f The desired height is found from the guidance. Writing
second order error dynamics for h, where error = h − h∗
(h f − hc )kx = tanγ ∗ (25)
(ḧ − h¨∗ ) + kḣ (ḣ − ḣ∗ ) + kh (h − h∗ ) = 0 (31)
Touchdown condition: The flare trajectory should intersect
the ground at touchdown point. Replacing x = xtd and h∗ = 0 ḧ = h¨∗ − kḣ (ḣ − ḣ∗ ) − kh (h − h∗ ) (32)
in (23) we get,
Under the assumption that the velocities do not change much
0 = hc + (h f − hc )e−kx (xtd −x f ) (26) in one time step, we differentiate ḣ equation (10) in order to
get ḧ. We can write symbolically
Sink rate at touchdown: The descent rate at touchdown
should be equal to specified sink rate. Differentiating (23) ḧ = ah + bh φ̇ + ch θ̇ (33)
and evaluating at x = xtd . Putting ḣ∗ = ḣtd
∗ , where ḣ∗ is the
td where ah , bh , ch are defined as
desired sink rate at touchdown. Here we assume the forward
velocity at touchdown is nearly equal to air velocity. ah , U̇sinθ − (V̇ sinφ + Ẇ cosφ )cosθ
bh , −(V cosφ −W sinφ )cosθ
ḣt∗ = −(h f − hc )kx ẋtd e−kx (xtd −x f ) (27)
ch , Ucosθ + (V sinφ +W cosφ )sinθ
Now we can solve the (24-27) for the four unknowns of Substituting back equation (33) in (32)
the landing trajectory. The solution will ensure a smooth
transition from glidelsope to flare path. We also have the ah + bh φ̇ + ch θ̇ = h¨∗ − kḣ (ḣ − ḣ∗ ) − kh (h − h∗ ) (34)
direct control over the touchdown point and sink rate at
Now separating the terms containing Q, in φ̇ and θ̇ from
touchdown. Which can be the design parameters and tuned
equation (7) and (8) respectively
as per the need. Now the goal of control design is to closely
track the desired landing trajectory. φ̇ = fφ + gφ Q (35)
θ̇ = fθ + gθ Q (36)
IV. C ONTROL D ESIGN
where, fφ , P + Rcosφ tanθ gφ , sinφ tanθ
Dynamic Inversion is a promising technique for nonlinear fθ , −Rsinφ gθ , cosφ
control design based on theory of feedback linearization.
Substituting (35) and (36) in (34) than rearranging
In dynamic inversion the nonlinearities are canceled by
feedback so that linear control system theory may be applied. h¨∗ − kḣ (ḣ − ḣ∗ ) − kh (h − h∗ ) − (ah + bh fφ + ch fθ )
This is achieved by enforcing stable error dynamics so that Q∗ = (37)
(bh gφ + ch gθ )
error goes towards zero by time hence achieve the tracking.
Here the control is divide into outer and inner loop. The 3) Bank angle control: To control bank angle roll rate
outer loop transforms the guidance commands into the body command is generated from the error in roll angle. Where
rates. Further inner loop generates the control required for desired roll angle is zero for level flight and in the case of a
tacking the desired body rates commanded by outer loop. turn it is found from the coordinated turn constraint. Writing
first order error dynamics for φ , where error = φ − φ ∗
A. Outer Loop Control (φ̇ − φ˙∗ ) + kφ (φ − φ ∗ ) = 0 (38)
1) Heading control: The heading is controlled during
φ̇ = φ˙∗ − kφ (φ − φ ∗ ) (39)
landing by generating yaw rate command from the error in
heading angle. The desired heading is found from the lateral separating terms containing P from φ̇ equation (7) and
deviation from runway. Writing first order error dynamics rearranging above equation we get,
for ψ , where error = ψ − ψ ∗
P∗ = φ˙∗ − kφ (φ − φ ∗ ) − (Qsinφ + Rcosφ )tanθ (40)
(ψ̇ − ψ˙∗ ) + kψ (ψ − ψ ∗ ) = 0 (28)
B. Inner Loop Control
1) Body rate control: To achieve successful landing the
ψ̇ = ψ˙∗ − kψ (ψ − ψ ∗ ) (29) control should be calculated such that it tracks the body
angular rates desired by the outer guidance loop. We can
Separating terms containing R from ψ̇ equation (9) and write objective as [P Q R]T −→ [P∗ Q∗ R∗ ]T . Let error
rearranging above equation we get, be e = [(P − P∗ ) (Q − Q∗ ) (R − R∗ )]T . Writing first order
error dynamics
R∗ = secφ cosθ (ψ˙∗ − kψ (ψ − ψ ∗ )) − Qtanφ (30)

2412
 C. Control structure
Ṗ − P˙∗
    
kP 0 0 P − P∗ Fig. 4 is the schematic diagram of the overall control
 Q̇ − Q̇∗  +  0 kQ 0   Q − Q∗  = 0 structure. The velocity loop is not shown in figure. Note that
Ṙ − R˙∗ 0 0 kR R − R∗ desire height (h∗ ), desires heading angle (ψ ∗ ) and desired
(41) roll angle (φ ∗ ) are coming from guidance.
P˙∗ − kP (P − P∗ )
   
Ṗ
 Q̇  =  Q̇∗ − kQ (Q − Q∗ ) 
Ṙ R˙∗ − kR (R − R∗ )
separating the state and control terms in Ṗ, Q̇, Ṙ from
equation (4-6) and rearranging above equation we can write
fR + gRUc = bR (42)
carrying out the necessary algebra, the solution is
Uc = g−1
R (bR − f R ) (43)
where, Uc = [δ a δ e δ r]T and other terms are defined as
follows
 
c1 RQ + c2 PQ + c3 Lax + c4 Nax
fR ,  c5 PR + c6 (P2 − R2 ) + c7 (Max − Mt ) 
c8 PQ − c2 RQ + c4 Lax + c9 Nax
  Fig. 4. Control structure
c3 Lau 0 c4 Nau
gR ,  0 c7 Mau 0 
c4 Lau 0 c9 Nau V. S IMULATION R ESULTS
P˙ − kP (P − P∗ )
 
∗ To evaluate the algorithm various cases have been simu-
bR ,  Q̇∗ − kQ (Q − Q∗ )  lated with different initial conditions of which two cases are
R˙∗ − kR (R − R∗ ) presented here. The simulation results are given along with
where, the result in tabulated form.

Lax , q̄Sb[Clβ (α ) β +ClP (α ) P +ClR (α ) R] A. Landing results

Max , q̄Sc[Cm0 +Cmα (α ) α +Cmβ (α , β ) β +CmQ (α ) Q] The complete landing sequence of approach, glideslope
and flare has been simulated. Two cases with different initial
Nax , q̄Sb[Cnβ (α ) β +CnP (α ) P +CnR (α ) R]
conditions are considered. One such case (Case 2) is also
and, considered in which the UAV has to loose height during
Lau , q̄SbClδ a ; Mau , q̄ScCmδ e ; Nau , q̄SbCnδ r approach. Fig. 5 shows the trajectory followed by the UAV
during landing. Initial part of the trajectory is approach where
2) Velocity control: The forward velocity is maintained it is aligning with runway. Next part is descent and flare
constant during landing by controlling the thrust through which are in longitudinal plane.
throttle. We can write error in velocity as (U − U ∗ ), where
U ∗ is the velocity which has to be maintained during landing.
Enforcing the first order error dynamics.
(U̇ − U̇ ∗ ) + kU (U −U ∗ ) = 0 (38) Case 2
50
∗ ∗ Case 1
U̇ = U̇ − kU (U −U ) (39) 40
Altitude (m)

Separating the state and control terms in U̇ from (1) and 30

rearranging 20

fU + gU σt = bU (40) 10

The control solution is 0

σt = gU−1 (bU − fU ) (41)

100
where, 0
50
fU , RV − QW − gsinθ + Xa −500
Tmax −1000
0
gU , Sideward distance (m) −1500
m Forward distance (m)
bU , U̇ ∗ − kU (U −U ∗ ) Fig. 5. Trajectory in 3-D during landing

2413
 TABLE II
Fig. 6 shows the landing trajectory in horizontal plane.
L ANDING RESULTS
Where as Fig. 7 shows the landing trajectory in vertical or
(x0 , y0 , h0 ) ψ0 (xg , yg ) xtd ḣtd
longitudinal plane. The Fig. 8 shows the control plots for m deg m m m/s
landing. It is observed that in initial part of landing there are Case 1 (-1500, 10, 50) 120o (-1000, 50) 48.76 -0.1001
some oscillations in control values this is because in initial Case 2 (-1500, 50, 60) 45o (-1100, 50) 48.97 -0.1001
segment the lateral and longitudinal both the maneuvers are
taking place.
120 VI. C ONCLUSIONS
Case 1
Case 2 In this paper a guidance and nonlinear control design
100
for automatic landing has been demonstrated. The simu-
80 lation results are promising and show the robustness of
the algorithm for different initial conditions. The glideslope
60
and flare path parameters are calculated online, such that it
y (m)

40
achieves the touchdown point. The sink rate at touchdown
remains within specified bounds. It is also ensured that
20 the transition from the glideslope to flare is smooth. The
dynamic inversion technique is used for nonlinear control
0
design. The underlying assumption of dynamic inversion is
−20 that we have a accurate knowledge of the dynamic model
−1500 −1000 −500 0
x (m) and true estimate of the states. However these problems can
Fig. 6. Trajectory in x-y plane during landing be addressed by augmenting dynamic inversion with neuro-
adaptive technique and using extended kalman filter for state
70 estimation. Further algorithm needs to be tested in presence
Case 1
Case 2 of wind shear and wind gust.
60
VII. ACKNOWLEDGEMENT
50
The authors would like to acknowledge the support
40 of UK-India Education Research Initiative for the project
h (m)

(CSSP/UKIE002) under which this research was funded.

30
R EFERENCES
20
[1] B. Prasad and S. Pradeep, ”Automatic Landing System Design us-
ing Feedback Linearization Method”, Proceedings of AIAA Infotech
10 Aerospace Conference and Exhibit, 2007, Rohnert Park, California.
[2] P. Rosa, C. Silvestre, D. Cabenchias and R. Cunha, ”Autolanding
0 Controller for a Fixed Wing Unmanned Air Vehicle”, AIAA Guidance
−1500 −1000 −500 0 Navigation and Control Conference and Exhibit, 2007, Hilton head,
x (m)
South Carolina.
Fig. 7. Trajectory in x-h plane during landing [3] R. Wang and Y. Shen, ”Flying-Wing UAV Landing Control and Sim-
ulation Based on Mixed H2 /H∞ ”, Proceedings of IEEE International
1 5 Conference on Mechatronics and Automation, 2007, Harbin, China.
Case 1
0.8 Case 2 0 [4] J. S. Shamma and M. Athans, ”Analysis of Gain Scheduled Control
for Nonlinear Plants”, IEEE Transactions On Automatic Control, Vol.
δt ( 0 − 1 )

−5
δe (deg)

0.6
−10 35, No. 8, 1990
0.4
−15
[5] D. Enns, D. Bugajski, R. Hendrick and G. Stein, ”Dynamic inversion:
0.2 −20
an evolving methodology for flight control design”, International
−25
Journal of Control, Vol. 59, Issue 1, 1994 , pages 71 - 91.
0
−1500 −1000 −500 0 −1500 −1000 −500 0 [6] K. S. Kumar and J.Shanmugam, ”Design of Blending function for
x (m) x (m)
landing phase of an Unmanned Air Vehicle” Proceedings of AIAA
15 15
Infotech Aerospace Conference and Exhibit, 2007, Rohnert Park,
10 10
California.
5 5
[7] V. Surendra nath, S. P. Govindaraju, M. S. Bhat and C. S. N.
δa (deg)

δr (deg)

0 0
Rao, ”Configuration Development of All Electric Mini Airplane”,
ADE/DRDO Project, 2004, Project Ref. No: ADEO/MAE/VSU/001,
−5 −5
Department of Aerospace Engineering, Indian Institute of Science,
−10 −10
Bangalore.
−15 −15
−1500 −1000 −500 0 −1500 −1000 −500 0 [8] E. Morelli, ”Nonlinear Aerodynamic Modelling Using Multivariate
x (m) x (m) Orthogonal Functions”, Proceedings of American Control Conference,
1998, Philadelphia, Pennsylvania.
Fig. 8. Throttle, elevator, aileron and rudder controls [9] B. Stevens and F. Lewis, ”Aircraft Control and Simulation 2nd
Edition”, J. Wiley & Sons, 2003.
The numerical results are given in Table II. It can be [10] M. Drella and H. Youngren, ”Athena Vortex Lattice Analysis 3.26
User Primer”, http://web.mit.edu/drela/Public/web/avl/, 2006.
observed that the desired touchdown point (50m) and desired
sink rate (−0.1m/s) at touchdown are achieved successfully.

2414