Vous êtes sur la page 1sur 6

Automatic Aircraft Landing Control Using Nonlinear Energy Method

Lim Jen Nee Jones, Rini Akmeliawati and Chee Pin Tan
AbstractThis paper presents the integration of longitudinal
and lateral dynamics controller designed using Nonlinear En-
ergy Method (NEM) for a research civil aircraft. This controller
ensures the core attitudes of the aircraft such as roll, pitch, yaw
and airspeed follows their reference value determined by the
given ight trajectory for a landing approach. The NEM is
supplemented with a positional tracker on the outer loop to
facilitate better positional tracking. The robustness and distur-
bance rejection of the proposed controller is tested in different
ight conditions. Simulation result shows that the NEM is stable
under scenarios such as a one-sided engine failure, a 90-degree
turn, glideslopes and a wind shear. Positional tracking is also
achieved in both normal and extreme ight conditions.
I. INTRODUCTION
Air travel has become a very common means transporta-
tion around the world. As more ying hours are required,
modern aircrafts rely heavily upon automatic ight control
systems. Control systems used on the aircraft has to be reli-
able and robust to undesirable conditions such as breakdowns
and wind disturbances.
Passivity-based control (PBC) theories form the underly-
ing approach to Nonlinear Energy Method (NEM), which
was known to be able to reject disturbance and is robust
to parametric uncertainties. Recent development on PBC
theories can be found in [4] and [5], which uses hamiltonian
dynamics. The PBC has also become increasingly popular
in aerospace application such as [6] and [10]. An automatic
ight control system using NEM was rst proposed by
[1], [2] and [3] to control the longitudinal dynamics of
Research Civil Aircraft Model (RCAM). However, the lateral
dynamics of the aircraft has to be regulated as well, in order
for the aircraft to be maneuverable. Hence, a similar NEM
was adopted to control the lateral dynamics in [8] and [9].
In this paper, we design an integrated longitudinal and
lateral control, whereby the design is done in one go using
NEM and nonlinear dynamics of the aircraft. Thus, the
organization of this paper is as follows, in Section II we
present the energy-based modeling of the aircraft. Section
III presents the NEM. Section IV presents the controller
designed using NEM for the aircraft automatic landing
system. Section V presents simulation results of the closed-
loop system, and Section VI concludes the paper.
L. J. N. Jones & Chee Pin Tan are with School of
Engineering, Monash University, 46150 Bandar Sunway, Selangor,
Malaysia lim.jen.nee@eng.monash.edu.my &
tan.chee.pin@eng.monash.edu.my
R. Akmeliawati is with Kulliyyah of Engineering, International
Islamic University Malaysia, 53100 Kuala Lumpur, Malaysia
rakmelia@iiu.edu.my
II. ENERGY-BASED MODELING OF AIRCRAFT
DYNAMICS
The Research Civil Aircraft Model (RCAM) is used in
this research. This twin-engine nonlinear aircraft model with
6 degrees of freedom is developed by the Group for Aeronau-
tical Research and Technology in Europe (GARTEUR) [7].
The full description of the mathematical model of RCAM
can be found in [11]. This aircraft can be classied as
an Euler-Langrangian (EL) system. Thus, the mathematical
model of this aircraft is derived using the Euler-Lagrange
equation, followed by a detailed description of the aircraft
mathematical model.
A. The Euler-Lagrangian Equation
A dynamical system can be dened by an EL equation
(1)[12] with n degrees of freedom, with a generalized coor-
dinate, q R
n
and external forces, Q R
n
,
d
dt
_
L
q(q, q)
_

L
q
(q, q) = Q, (1)
where the Lagrangian function is
L(q, q)

= T (q, q) V(q) (2)
and T (q, q) is the kinetic energy function, which we assume
to be of the form
T (q, q) =
1
2
q
T
D q. (3)
D R
nn
is the generalized inertia matrix that satises
D = D
T
> 0 and V(q) is the potential energy function
which is assumed to be bounded from below, that is, there
exists a c R
n
such that V(q) c for all q R
n
.
Dissipative forces are of the form
F
q
( q), where F( q) is a
Rayleigh dissipation function such as aerodynamic forces of
the aircraft, satisfying
q
T
F
q
( q) 0. (4)
The external forces are dened by
Q =
F
q
( q) + M
u
u, , (5)
where M
u
is the input matrix and u is the controller input.
In summary, the EL equation of motion
d
dt
_
L
q(q, q)
_

L
q
(q, q) +
L
q
( q) = M
u
u, (6)
2010 IEEE International Conference on Control Applications
Part of 2010 IEEE Multi-Conference on Systems and Control
Yokohama, Japan, September 8-10, 2010
978-1-4244-5363-4/10/$26.00 2010 IEEE 2142
with (2), (3) and (5) denes an EL system. Finally, the system
equation for the rotational dynamics is derived in (7).
D q +
F
q
= M
u
u (7)
B. The Research Civil Aircraft Model (RCAM)
The RCAM can be written as an EL equation in (7) as a
function of aircraft total mass, m:
D =
_
_
I
x
0 0
0 I
y
0
0 0 I
z
_
_
=
_
_
40.07m 0 0
0 64m 0
0 0 99.92m
_
_
(8)
I
x
, I
y
and I
z
are the aircraft inertia relative to the vehicle-
carried vertical frame in the x, y and z direction respectively.
This model is derived in such a way that the states of the
aircraft dynamics are based on the vehicle-carried vertical
frame where its orientation follows the earth reference frame.
Generalized coordinates are described by q =
[ ]
T
and their respective rotational velocities,
q = [


]
T
where the Euler angles are represented
by roll, , pitch, and yaw, . The control surfaces
are the ailerons, elevator and rudder, represented by
u = [
a

e

r
]
T
. These inputs correspond primarily
to dynamics in q and q but are decoupled using the input
matrix, M
u
as
M
u
=
_
_
M
,a
0 M
,r
0 M
,e
0
0 0 M
,r
_
_
. (9)
The rows of M
u
represent the 3 fast dynamics of interest
while the columns of M
u
relate those dynamics to each of
the controller input. Some components of M
u
were found
to be zeros as some dynamics are not inuenced directly
by certain control surfaces. Subsequently, dissipative forces
caused by aerodynamic forces such as Lift, Drag, Thrust,
and Side Force of the aircraft are described by
F
q
=
_
F

_
T
. (10)
Components of both the input matrix and dissipative forces
can be derived from the RCAM.
III. NONLINEAR ENERGY METHOD (NEM)
A particular system governed by the NEM has to be
modeled with an EL equation. The control laws are then
developed based on Lyapunov and Invariant Set theorems.
The NEM is elaborated in the following proposition.
Proposition 1. Consider the aircraft equations of motion
as dened by (7), with the control input in (6) is dened as
M
u
u =
_
D q
ref
K
d
(

q + q) +
F
q
( q)
_
, (11)
where K
d
= K
T
d
> 0, R
+
are design parameters, then
the system (7) is stable in Lyapunov sense.
Proof: Using the storage function
1
1
() denotes the error between a state variable and its desired reference
value
S(q, q) =
1
2

q
T
D

q +
1
2
q
T
K
d
q, (12)
the derivative of S(q, q) with respect to time is

S(q, q) =

q
T
D

q +

q
T
K
d
q. (13)
Substituting (11) into (7) yields
D

q + K
d

q + K
d
q = 0. (14)
Subsequently, we obtain

S(q, q) =

q
T
(K
d

q K
d
q) +

q
T
K
d
q, (15)
by substituting (14) into (13), which gives us

S(q, q) =

q
T
K
d

q 0. (16)
Hence, Lyapunov stability [13] is achieved.
Let

q = 0 be the set of all points within a bounded region
where

S(q, q) = 0 and q = 0 be the largest invariant set
in

q = 0. Then by using the Invariant Set Theorem, every
solution for q originating in converges to q = 0 as t .
When

q = 0, the system equation in (14) can be rewritten
as

q =
K
d
q
D
0. This implies that the acceleration is
always decreasing. Furthermore,

q = 0 if and only if q = 0.
Therefore, the system will always converge to q = 0 and
will remain at [ q = 0,

q = 0]. Since

S(q, q) is unbounded as
q , the region is unbounded. Thus, the equilibrium
point at [ q = 0,

q = 0] is globally asymptotically stable and
the errors q and

q will converge to zero.
IV. NEM CONTROLLER DESIGN
The NEM controller is designed for the control surfaces
that regulates airspeed, roll, pitch and yaw in the inner-loop
of the controller. A positional tracker acts as the outer-loop
controller. Therefore, the reference values for airspeed, roll,
pitch and yaw are determined by the positional tracker.
A. Control surfaces
The RCAM has 5 control surfaces. They are the ailerons,
elevator, rudder and 2 throttle inputs. Since the RCAM is a
twin-engine aircraft, each engine is controlled by one throttle
input. The throttle input,
th
in (17)
2
is taken from the basic
controller designed in [3],

th
=
+ g sin( ) +

V
A
ref
K
dV
(V
A
V
A
ref
)
19.62 cos()
,
(17)
where the angle of attack is dened by , airspeed denoted
by V
A
and
=
25.47V
2
A
+ 337.2V
2
A

2
+ 80.2V
2
A

m
. (18)
Design constant, K
dV
= 1 is applied to regulate the airspeed.
No modication was done on the original throttle controller
2
()
ref
denotes the desired reference value
2143
TABLE I
DESIGN CONSTANTS
K
dx
= 1961827 x = 0.283
K
dy
= 3317760 y = 0.833
K
dz
= 3453235 z = 0.556
designed by Akmeliawati [3] since its control structure is also
based on the NEM. Both engines utilizes the same controller
except that the left engine is set to 0.5
o
with rst order
system dynamics given by a rst order lter represented by
the transfer function,
1
1 + 3.3s
, (19)
when a one-sided engine failure is simulated. Further elab-
oration on how the Passivity-Based Control technique was
implemented to derive the throttle input can be found in [1],
[2] and [3].
With reference to (8)-(11) and Table I, the attitude con-
trollers of the aileron, elevator and rudder are derived in (20),
(21) and (22) based on Proposition 1.

a
=
I
x

ref
+
F

K
dx
(

+
x

) M
,r

r
M
,a
(20)

e
=
I
y

ref
+
F

K
dy
(

+
y

)
M
,e
(21)

r
=
I
z

ref
+
F

K
dz
(

+
y

)
M
,r
(22)
B. Positional tracker
As the NEM controller regulates the aircrafts attitudes
(roll, pitch, yaw and airspeed), a positional tracker deter-
mines these attitudes based on positional feedback. This
positional tracker is placed before the NEM as an outer-
loop controller as it computes the necessary desired atti-
tudes. Simple classical control techniques were good enough
to eliminate positional errors by complementing the NEM
controller. A Proportional-Integral (PI) controller is included
as part of the positional tracker for pitch and yaw reference
angles to ensure altitude and lateral tracking respectively. The
breakdown of the positional trackers and how the aircraft
attitudes are determined are elaborated accordingly.
1) Roll angle: The desired roll angle is determined based
on its relationship with the desired track angle rate,
ref
and
airspeed using (23), as follows:

ref
= tan
1
_
V
A
ref

ref
g
_
(23)
The
ref
signal passes through a lter (19) before it is being
fed to the NEM controller. This lter improves passengers
comfort by preventing an abrupt step input in the roll
command. Both desired roll rate and roll acceleration are
set to zero.
2) Pitch angle: As for the desired pitch angle described
by (24), it is determined by desired ight path angle,
ref
and . The positions are dened in the aircrafts body-xed
reference frame denoted by x
b
, y
b
and z
b
while those in the
aircrafts vehicle-carried vertical frame are denoted by x
v
,
y
v
and z
v
. The desired pitch angle is dened by

ref
=
ref
+ + K
pz
z
b
+ K
iz
_
z
b
. (24)
where
ref
= tan
1
_
zv
ref
_
x
2
v
ref
+ y
2
v
ref
_
, K
pz
= 0.001
and K
iz
= 0.00001. The desired pitch rate and pitch
acceleration are set to zero.
3) Yaw angle: A similar positional tracker is implemented
as part of the function to determine the desired yaw angle
in (25). The desired yaw angle is dened by

ref
=
ref
+ K
py
y
b
+ K
iy
_
y
b
, (25)
where the desired track angle,
ref
= tan
1
_
yv
ref
xv
ref
_
,
K
pz
= 0.001, K
iz
= 0.0001 and the sideslip angle denoted
by . Yaw acceleration is set to zero. However, yaw rate is
set equal to
ref
whenever the aircraft makes a turn.
4) Airspeed: Finally, the airspeed is then slightly modied
by adding a Proportional (P) controller to ensure that x
b
can
be tracked, as shown by
V
A
ref
= V
A
ref0
+ K
px
x
b
, (26)
where K
px
= 0.001 and V
A
ref0
is typically 80 m/s on a
normal ight condition. All P and PI gains chosen above
are very small but necessary to ensure the difference in time
scale.
V. EVALUATION OF THE NEM CONTROLLER
This section explains how the NEM controller is evaluated
for its stability and robustness. The evaluation includes a
description of the ight trajectory, ight conditions and
simulation results.
A. Flight test trajectory
The ight test trajectory runs for 510 seconds of simulation
time with desired constant airspeed of 80 m/s. There are
4 ight scenarios to assess the controller in 10 different
ight conditions. The aircraft starts off by heading toward
the positive direction of the earth reference y-axis, as it
prepares for landing. The rst scenario consists of a one-
sided engine failure located on the right wing (starboard)
and this engine will be restored in order to complete the
rest of the trajectory. The second challenge requires the
aircraft to make a 90
o
right turn toward positive x-axis of
earth reference frame, with a standard turning rate of 3
o
per
second. As the plane continues to y in the direction of
positive x-axis, the pitch angle is rotated from 1.6272
o
to
4.3728
o
, after which, it rotates to 1.3728
o
degrees. The
pitch angles were calculated from the ight path angles of
6
o
and 3
o
respectively. The specications of the ight test
trajectory are summarized in Table II. The ight trajectory
2144
TABLE II
SEQUENCE OF FLIGHT SCENARIO FOR RCAM
Simulation Time (s) Description of Flight Scenario
0 Start of simulation with straight ight path
40 One-sided engine failure begins
120 One-sided engine failure ends
220 90
o
turn begins
250 90
o
turn ends
315 Pitch angle begins to rotate to 4.3728
o
335 Pitch angle stops rotating
360 Pitch angle begins to rotate to 1.3728
o
370 Pitch angle stops rotating
395 Wind shear begins
495 Wind shear ends
510 End of simulation
Fig. 1. Wind velocities during wind shear [7]
described follows the evaluation procedure of Segment I to
IV of the RCAM mission and scenario found in [7].
The wind shear model is active from 395 s to 495 s in the
earth-xed x-axis (WXE) and z-axis (WZE). As the aircraft
completes the trajectory, it will be experiencing a headwind
going up to about WXE = -7 m/s, then wind speed will
change to a tailwind of about WXE = 7 m/s, combined with
a downdraught of about WZE = 8 m/s as seen in Fig. 1.
B. Flight condition
A normal ight condition for RCAM is dened by pa-
rameters such as the aircraft total mass, positions of the
center-of-gravity (CoG) in the measurement reference frame
and the presence of wind disturbances. Ten different ight
conditions were devised, which ranges from a normal ight
operating condition to an extreme ight condition. The
range of possible parameter choices in RCAM as given in [7]
are summarized in Table III. The aircraft mean aerodynamic
chord of 6.6 m is denoted by c.
Different combinations of these parameters were chosen
to make up 10 ight conditions for the purpose of testing
the robustness and stability of the controller. These ight
conditions are described in details found in Table IV wiht
reference to the order of parameters specied in Table III.
Simulations 4, 5, 9 and 10 include a xed heading constant
wind of -10 m/s that is directed along the negative earth-xed
x-axis. This constant wind has no vertical component. It acts
as a crosswind before the 90
o
turn, becomes a headwind after
the turn and is slowly reduced to zero between 250 s - 395
TABLE IV
DESCRIPTION OF FLIGHT CONDITIONS
Simulation No. Flight Condition Parameters
1 (Normal) {120 000, 0.23 c, 0, 0, 0, Off, 80}
2 {100 000, 0.23 c, 0, 0, 0, Off, 80}
3 {150 000, 0.23 c, 0, 0, 0, Off, 80}
4 {120 000, 0.23 c, 0, 0, -10, Off, 80}
5 {120 000, 0.23 c, 0, 0, -10, On, 80}
6 {120 000, 0.31 c, 0, 0.21 c, 0, Off, 80}
7 {120 000, 0.31 c, 0.03 c, 0.21 c, 0, Off, 80}
8 {120 000, 0.23 c, 0, 0, 0, Off, 90 70}
9 {150 000, 0.31 c, 0, 0.21 c, -10, On, 80}
10 (Extreme) {150 000, 0.31 c, 0.03 c, 0.21 c, -10, On, 90 70}
s, just before wind shear begins.
A Dryden turbulence eld of 305 m scale length and
0.08 m/s amplitude is assumed to be active in Simulations
5, 9 and 10. This recommended amplitude is only 5% of
the amplitude for moderate conditions to prevent the effect
of turbulence on lateral and longitudinal accelerations from
overruling other effects that we are interested in.
As for the reference airspeed, all simulations run at 80 m/s
except for Simulations 8 and 10. The airspeed in these two
simulations start off with 90 m/s and decelerates to 80 m/s at
200 s. The aircraft basically keeps a constant airspeed of 80
m/s until 390 s and continues to decelerate as it goes through
a wind shear until the airspeed reaches 70 m/s a the end of the
simulation time. Although the recommended airspeed is 80
m/s for all ight conditions, this variable reference airspeed
was introduced to simulate deceleration of the aircraft, which
is quite common in the landing approach. It is also interesting
to observe how well the controller tolerates to a variable
airspeed command. The airspeed requirement is relaxed in
order to cope well with the engine failure and severe wind
disturbances.
C. Simulation Results
The performance of the NEM controller with a positional
tracker under the 4 ight scenarios of a one-sided engine
failure, a 90
o
turn, variable glideslope and wind shear will be
presented in this subsection. The responses of the airspeed
and throttle inputs are presented as a prelude to the main
results. The NEM controller designed by [3] adapts well to
the NEM controller designed to control rotational dynamics
in this paper, as shown in Fig. 3. In a normal ight condition,
it is evident that the throttle input on the right engine puts in
a lot more effort without reaching its saturation limits when
the left engine fails. The airspeed regulator continues to track
the desired airspeed command close to 80 m/s in spite of a
90
o
turn, glideslope and wind shear.
The main results focuses on positional tracking and the
performance of the NEM controller designed for rotational
dynamics of the aircraft. Fig. 2 shows a comparison of how
the controller tracks the recommended trajectory in a normal
and extreme ight condition, as described in Table IV.
The tracking performance can be shown in breakdowns of
4 ight scenarios mentioned earlier as 2-dimensional plots
in more appropriate views listed in Fig. 4 and Fig. 5. Since
2145
TABLE III
POSSIBLE PARAMETER CHOICES IN RCAM
Parameters Bounds or Settings Normal Flight
m 100 000 kg m 150 000 kg 120 000 kg
X position of CoG, Xcg 0.15 c Xcg 0.31 c 0.23 c m
Y position of CoG, Y cg -0.03 c Y cg 0.03 c 0.00 c m
Z position of CoG, Zcg 0.00 c Zcg 0.21 c 0.00 c m
Fixed heading constant wind 0 or -10 m/s 0
Dryden turbulence eld On or Off Off
V
A
70 m/s V
A
90 m/s 80 m/s
0
1
2
x 10
4
1
0
1
2
x 10
4
0
500
1000

xposition (XE) [m]
(a) Normal Flight Condition
yposition (YE) [m]

a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
reference
actual
90degree
turn
Glideslope
Wind shear
Onesided
Engine Failure
0
1
2
x 10
4
1
0
1
2
x 10
4
0
500
1000

xposition (XE) [m]
(b) Extreme Flight Condition
yposition (YE) [m]

a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
90degree
turn
Glideslope
Wind shear
Onesided
Engine Failure
Fig. 2. Positional Tracking of Recommended Flight Trajectory
the reference roll, pitch and yaw angles are determined by
the positional tracker, the behavior and tracking of these
Euler angles are different in each situation. Tracking of these
angles are presented in Fig. 6 and Fig. 7 accordingly. The
inputs of the aileron, elevator, rudder and throttle act upon the
desired values of roll, pitch, yaw and airspeed respectively.
The most obvious difference found in Fig. 4 as compared
to Fig. 5 are the random deviations that are most likely
due to wind turbulence. However, the controller is able to
pull the position of the aircraft back to its trajectory after a
new scenario was introduced at each stage. The wind shear
effects are similarly seen both in the normal and extreme
ight conditions. The reference positional trajectories gener-
ated between both conditions are also different because the
airspeed in without changing the total simulation time.
A deviation of almost 100 m is found in Fig. 5 just
before the engine fails, as the nonlinear aircraft model is
not trimmed to the existing change in parameters and wind
conditions. Nevertheless, the controller reduces the effects
of this sudden side wind rapidly before one of the engine
fails. Most importantly, we see that the controller is still
stable and ensures that the aircraft reach its nal destination
at the end of simulation. The aircraft continues to track the
reference position in all 4 scenarios, even in the extreme
ight condition, although more deviation is seen in the 2-
dimensional plots.
Fig. 6 and Fig. 7 shows how the reference values of
0 100 200 300 400 500
0
5
10
(b) Left Engine
t(s)
t
h
r
o
t
t
l
e
1

(
d
e
g
.
)


0 100 200 300 400 500
0
5
10
(c) Right Engine
t(s)
t
h
r
o
t
t
l
e
2

(
d
e
g
.
)


0 100 200 300 400 500
70
72
74
76
78
80
82
84
86
88
90
(a) Airspeed
t(s)
V
A

(
m
/
s
)


actual
reference
EF
EF
TU
TU
GL
GL
WS
WS
Glideslope (GL)
90degree turn (TU)
Onesided Engine Failure (EF)
Wind shear
(WS)
Fig. 3. Airspeed Response and Throttle Inputs
2 1.5 1 0.5
x 10
4
200
100
0
100
200
yposition (YE) [km]
x

d
e
v
i
a
t
i
o
n

[
m
]
The effect of engine failure


2.4 2.2 2 1.8 1.6
x 10
4
4000
2000
0
2000
xposition (XE) [km]
y

p
o
s
i
t
i
o
n

(

Y
E
)

[
k
m
]
Plan view of 90degree turn
1.6 1.4 1.2 1
x 10
4
600
800
1000
xposition (XE) [km]
a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
Side view of 6 and 3 degree glideslope
10000 5000 0
0
200
400
600
xposition (XE) [km]
a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
Side view of the final approach with wind shear
reference
actual
Fig. 4. Positional Tracking in 4 ight scenarios for Normal Flight Condition
2 1.5 1 0.5
x 10
4
200
100
0
100
200
yposition (YE) [km]
x

d
e
v
i
a
t
i
o
n

[
m
]
The effect of engine failure


2.4 2.2 2 1.8 1.6
x 10
4
4000
2000
0
2000
xposition (XE) [km]
y

p
o
s
i
t
i
o
n

(

Y
E
)

[
k
m
]
Plan view of 90degree turn
1.6 1.4 1.2 1
x 10
4
600
800
1000
xposition (XE) [km]
a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
Side view of 6 and 3 degree glideslope
10000 5000 0
0
200
400
600
xposition (XE) [km]
a
l
t
i
t
u
d
e

(

Z
E
)

[
m
]
Side view of the final approach with wind shear
reference
actual
Fig. 5. Positional Tracking in 4 ight scenarios for Extreme Flight
Condition
2146
0 50 100 150 200 250 300 350 400 450 500
20
0
20
40
(a) roll
t(s)
d
e
g
.


0 50 100 150 200 250 300 350 400 450 500
10
0
10
(b) pitch
t(s)
d
e
g
.
0 50 100 150 200 250 300 350 400 450 500
100
50
0
50
(c) yaw
t(s)
d
e
g
.


reference
actual
Wind shear
90degree turn
90degree turn
Onesided Engine Failure
Glideslope
90degree turn
Fig. 6. Tracking and behavior of Roll, Pitch and Yaw angles in Normal
Flight Condition
0 50 100 150 200 250 300 350 400 450 500
20
0
20
40
(a) roll
t(s)
d
e
g
.


0 50 100 150 200 250 300 350 400 450 500
10
0
10
20
(b) pitch
t(s)
d
e
g
.
0 50 100 150 200 250 300 350 400 450 500
200
100
0
100
(c) yaw
t(s)
d
e
g
.


reference
actual
90degree turn
90degree turn
90degree turn Glideslope
Wind shear
Fig. 7. Tracking and behavior of Roll, Pitch and Yaw angles in Extreme
Flight Condition
pitch and yaw angles could differ depending on the values
generated by the positional tracker. Roll commands for both
normal and extreme conditions are the same, since the
rolling angle is a direct function of the track angle rates.
The maximum roll angle should be limited to 30
o
based
on safety criteria, which in both normal and extreme case,
the roll angle adheres to its limits. Besides that, both pitch
and yaw angles tracks their command values very well.
A more detailed evaluation of these rotational dynamics
performances will be addressed in the form of root-mean-
square error (RMSE) values plotted in Fig. 8.
Results shown in Fig. 8 depict the performance of the
NEM controller on its own, without a direct inuence of
the positional tracker. The RMSE values for all the Euler
angles are particularly low regardless of how bad the ight
conditions are. If the angles are too big, the instantaneous
position of the aircraft will easily be far off from the desired
position with quite an immediate effect. We can see that
the roll angle has consistently low RMSE values except for
Simulation 3 whereby the aircraft total mass is changed to
150 000 kg. Interestingly, the RMSE of pitch angle in the
normal and extreme conditions are not very different. The
high aircraft total mass affects the performance yaw angle as
well. Besides mass variations, the combination of a constant
wind superimposed with turbulence that was introduced in
Simulations 5, 9 and 10 slightly affect the RMSE values of
pitch and yaw.
VI. CONCLUSION
A NEM controller was designed to track roll, pitch and
yaw commands of a twin-engine research civil aircraft. This
0
0.5
1
1.5
2
2.5
1 2 3 4 5 6 7 8 9 10
Flight Condition
R
M
S
E

v
a
l
u
e
s

(
d
e
g
.
)
RMSE of Roll Angle
RMSE of Pitch Angle
RMSE of YawAngle
Fig. 8. Root-mean-square errors (RMSE) of Roll, Pitch and Yaw Tracking
proposed NEM controller has been successfully integrated
with a similar NEM controller to regulate airspeed. Positional
tracking is achieved using a simple linear controller to
determine the required airspeed, roll, pitch and yaw angles.
The integrated proposed controller used a nonlinear aircraft
model. Thus, the nonlinearity in the dynamical equations
can be utilized to provide better control. The performance
of the NEM controller has been shown in this paper. Good
positional tracking is evident in normal as well as extreme
ight condition.
REFERENCES
[1] R. Akmeliawati and I. Mareels, Flight control systems using
passivity-based control - Disturbance rejection and robustness anal-
ysis, in AIAA Guidance, Navigation, and Control Conference and
Exhibit, vol. 2, 1999.
[2] R. Akmeliawati and I. Mareels, Nonlinear energy-based control
method for aircraft dynamics, in Proceedings of the 40th IEEE
Conference on Decision and Control, 2001, pp. 658-663.
[3] R. Akmeliawati and I. Mareels, Nonlinear energy-based control
method for aircraft automatic landing systems, to appear in IEEE
Transactions on Control Systems Technology, Sept. 2010.
[4] A. Astol and R. Ortega, Dynamic extension is unnecessary for
stabilization via interconnection and damping assignment passivity-
based control, Systems & Control Letters, vol. 58, 2009, pp. 133-135.
[5] F. Castanos and R. Ortega, Energy-balancing passivity-based control
is equivalent to dissipation and output variance, Systems & Control
Letters, vol. 58, 2009, pp. 553-560.
[6] J. Erdong and S. Zhaowei, Passivity-basd control for a exible
spacecraft in the presence of disturbances, International Journal of
Non-Linear Mechanics, vol. 45, 2010, pp. 348-356.
[7] FM(AG08), Robust Flight Control Design Challenge Problem For-
mulation and Manual: the Research Civil Aircraft Model (RCAM),
Technical Report GARTEUR/TP-088-3, GARTEUR, 1997.
[8] L. J. N. Jones and R. Akmeliawati and C. P. Tan, Roll and Yaw
Stabilisation using Nonlinear Energy Method, in Proceedings of the
9th International Conference on Control, Automation, Robotics and
Vision, 2006, pp. 803-806.
[9] L. J. N. Jones and R. Akmeliawati and C. P. Tan, Aircraft Automatic
Maneouvering System Using Energy-based Control Technique, in
Proceedings of the 17th World Congress of The International Fed-
eration of Automatic Control, 2008, pp. 1200-1205.
[10] R Kristiansen and P. J. Nicklasson and J. T. Gravdahl, Spacecraft
coordination control in 6DOF: Integrator backstepping vs passivity-
based control, Automatica, vol. 44, 2008, pp. 2896-2901.
[11] G. Looye and S. Bennani, Description and Analysis of the Research
Civil Aircraft Model (RCAM), Technical Report GARTEUR/TP-088-
27, GARTEUR, 1997.
[12] R. Ortega and A. Loria and P. J. Nicklasson and H. Sira-Ramirez,
Passivity-based Control of Euler-Lagrange Systems: Mechanical, Elec-
trical and Electromechanical Applications, Springer-Verlag London
Limited, 1998.
[13] J.E. Slotine and W. Li, Applied nonlinear control, Prentice Hall, 1991.
2147

Vous aimerez peut-être aussi