ARTICLE IN PRESS

Control Engineering Practice 13 (2005) 601–612

Control and simulation of arbitrary flight trajectory-tracking

T.S. Noa,*, B.M. Mina, R.H. Stoneb, K.C. Wongb
a
Department of Aerospace Engineering, Chonbuk National University, Deokjin Dong, Chonju 560-756, Republic of Korea
b
Department of Aerospace Engineering, University of Sydney, NSW 2006, Australia
Received 24 April 2003; received in revised form 6 May 2004; accepted 12 May 2004
Available online 17 June 2004

Abstract

A detailed description of a six-degree-of-freedom nonlinear flight simulation is presented and a method of generating the guidance
commands for controlling the aircraft trajectory in three-dimensional space is proposed. Tracking errors are expressed in terms of
zero effort miss vector, which is the expected miss distance at the future time between two flight vehicles, that is, one on the desired
trajectory and the other on the actual trajectory. This zero effort miss vector is used to form a Lyapunov function. Then a set of
guidance commands for the speed and flight direction control loops are selected in such a way that they will strictly decrease the
Lyapunov function. Thus obtained guidance commands are incorporated into the existing flight control systems and tested with
representative flight scenarios where the performance of trajectory tracking is considered to be important.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Nonlinear flight simulation; Flight path controllers; Trajectory-tracking control; Guidance laws; Zero effort miss; Lyapunov stability
theorem

1. Introduction this, the precise position control in the three-dimen-

sional space is necessary.
There are many situations where the flight vehicle There seems to be two extreme approaches in
should follow a predetermined path in a three-dimen- designing the flight trajectory-tracking controllers.
sional space. For example, the instrument landing Given a reference trajectory to follow, one completely
system (ILS) for a conventional, fixed-wing aircraft separates the guidance problem from the controller
requires a precise control of the positional error from design one. The other integrates both problems. It is
the prescribed glide and flare paths while keeping its typical that the guidance laws are usually designed
speed as scheduled (Shue & Agarwal, 1999; Fujimori, using a relatively simple model (Schultz, 1990; Heymann
Kurozumi, Nikiforuk, & Gupta, 2000). With the & Ben-Asher, 1997). They are directly used in the
advancements in electronic technology and precise subsequent tracking and stabilization control loops
navigation systems, fully or partially automated un- as input commands (Al-Hiddabi & McClamroch,
manned aerial vehicles (UAV) are being used and 2002). The trajectory-tracking performance can be
developed for various missions. Such UAVs are often further improved if both guidance and control pro-
required to follow a pre-designated trajectory (Boyle & blems are integrated and solved together (Jung & Hess,
Chamitof, 1999; Kaminer, Pascoal, Hallberg, & Silves- 1991; Marconi et al., 2002; Shue & Agarwal, 1999). Shue
tre, 1998; Wu, Engelen, Babuska, Chu, & Mulder, 2003) and Agarwal (1999) have applied mixed H2 =HN method
or operate in a restricted environment for its take-off in obtaining the command generator and tracking
and landing (Marconi, Isidori, & Serrani, 2002). For controller dedicated to the automatic landing of a
conventional fixed-wing aircraft. In Jung and Hess
(1991), the predicted output of the tracking error
*Corresponding author. Tel.: +82-632-702-470; fax: +82-632-702- is effectively used to reshape the command input so
471. that a zero tracking error and internal stability are
E-mail address: [email protected] (T.S. No). guaranteed.

0967-0661/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conengprac.2004.05.002
 ARTICLE IN PRESS
602 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

In other previous works, the trimmed flight conditions 2. Aircraft dynamics modeling
along the reference trajectory are used as the command
input to the tracking controllers (Kaminer et al., 1998; 2.1. Equations of motion
Boyle & Chamitof, 1999). For example, Boyle and
Chamitof (1999) used the quasi-trimmed flight condition Based on the assumption of the flat Earth and
in order to convert the given guidance commands into constant mass properties, the generic equations of
the commands for the inner loop to follow. These translation and rotational motion of a flight vehicle
commands are fed forward and augmented with the may be expressed as follows:
robust maneuver-tracking controller to account for any
 X
qV
uncertainty in the system modeling and disturbance. A m þxV ¼ F; ð1Þ
qt
similar approach was used in Kaminer et al. (1998),
where a series of tracking controllers were designed qH X
based on the linearized dynamics about the reference þxH ¼ M: ð2Þ
qt
trajectory.
Referring to Stevens and Lewis (1992) and following the
However, the previous results may not be readily
symbols and their definition used therein, Eqs. (1) and
applicable if the vehicle dynamic characteristics along
(2) are expanded, together with appropriate kinematic
the desired path are not fully known beforehand or the
equations, to write
pre-existing tracking controller cannot be easily mod-
Force equations:
ified for trajectory-tracking purposes. For example, the
flight trajectory of an aircraft is indirectly controlled by U’ ¼ RV  QW  g0 sin y þ Fx =m;
changing the speed and flight direction: flight path and V’ ¼ RU þ PW þ g0 sinf cos y þ Fy =m;
heading angles. Gibbens (1998) converted the position ’ ¼ QU  PV þ g0 cosf cos y þ Fz =m:
W ð3Þ
information of the reference trajectory in the fixed frame
into such as heading, rolling, and altitude commands. Kinematic equations:
This is because the existing vehicle control system is not ’
f ¼ P þ tan yðQ sin f þ R cos fÞ;
configured for direct control of its positional error.
Unless a dedicated controller for trajectory-tracking is y’ ¼ Q cos f  R sin f;
available, this will not necessarily lead to a precise match c’ ¼ Q sin f=cos y þ R cos f=cos y: ð4Þ
in position and velocity between the desired and actual
trajectories. Moment equations:
Therefore, the motivation behind this paper is to ’
P ¼ ðc1 R þ c2 PÞQ þ c3 L þ c4 N;
propose a simple method of generating the gui- Q’ ¼ c5 PR  c6 ðP2  R2 Þ þ c3 M; ð5Þ
dance commands for the trajectory-tracking problem
R’ ¼ ðc8 P  c2 RÞQ þ c4 L þ c9 N:
in three-dimensional space. The reference trajectory
is prescribed as a function of time, and it is assumed Navigation equations:
that the tracking controllers have already been de- p’ N ¼ U cos y cos c þ V ðcos f sin c þ sin f sin y cos cÞ
signed to accept the input commands in terms of speed
þ W ðsin f sin c þ cos f sin y cos cÞ;
and flight directions angles only. Since the reference
trajectory and the tracking controllers are designed p’ E ¼ U cos y sin c þ V ðcos f cos c þ sin f sin y sin cÞ
separately, the guidance scheme should accept the þ W ðsin f cos c þ cos f sin y sin cÞ;
tracking errors and transformed them into the in- ’
h ¼ U sin y  V sin f cos y  W cos f cos y: ð6Þ
put commands that are readily acceptable by the
controllers. Fig. 1 and Table 1 are referred to for the definitions of
For this purpose, the trajectory-tracking problem is variables and parameters used in the above equations. It
transformed into that of the missile–target intercept. should be noted that these equations describe the fully 6-
From the missile–target intercept geometry, the zero DOF nonlinear motion of an aircraft in three-dimen-
effort miss distance between the missile and the target is sional space. The complexity of the equations of motion
defined and used to form a Lyapunov function. Then, is further increased due to the nonlinearity of and
Lyapunov’s stability theorem is employed to obtain the coupling with the aerodynamic forces ðFx ; Fy ; Fz Þ and
guidance commands. This approach has been success- moments ðL; M; NÞ:
fully used for one- and two-dimensional problems (No,
Chong, & Rho, 2001a; No, Cochran, & Kim, 2001b). 2.2. Aerodynamic force and moment
Results obtained from the fully six-degree-of-freedom
(6-DOF) nonlinear simulation of representative flight Modeling fidelity of the aircraft dynamics is highly
scenarios are presented to demonstrate the applicability dependent on that of aerodynamic force and moment
of the proposed method. modeling. Aerodynamic forces in terms of lift, drag, and
 ARTICLE IN PRESS
T.S. No et al. / Control Engineering Practice 13 (2005) 601–612 603

YB YW Table 2
Aircraft physical properties
V
S ðwing areaÞ ¼ 1703:4ðft2 Þ; bðwing spanÞ ¼ 119:1ðftÞ; cðwing chordÞ ¼ 16:2ðftÞ
M , Q, θ
X  ðreference center of gravityÞ ¼ 0:25; X ðaircraft center of gravityÞ ¼ 0:31
c:g c:g

ie ðengine incidence angleÞ ¼ 2:0ðdegÞ

Ixx ¼ 515; 698:7 ðslug  ft2 Þ; Iyy ¼ 1; 044; 843:4ðslug  ft2 Þ
Iyy ¼ 1; 433; 408:8ðslug  ft2 Þ; Ixz ¼ 60636:7ðslug  ft2 Þ
tE ¼ tA ¼ tR ðelevator; aileron; rudder servo time constantÞ ¼ 0:05ðsecÞ
VT β
tt ðthrottle servo time constantÞ ¼ 0:1ðsecÞ
XW
α mg tEPL ðengine time constantÞ ¼ 0:5ðsecÞ

U N , R, ψ
L, P , φ
XB W
ZB

Fig. 1. Symbols used to denote the aircraft dynamics. Table 3

Non-dimensional aerodynamic force and moment coefficients

CLtotal ¼ CL ðM; df ; aÞ þ De CL ðM; de ; aÞ þ c=ð2VÞCLq ðM; aÞq

Table 1 CDtotal ¼ CD ðM; df ; aÞ þ De CD ðM; de ; aÞ
Definition of parameters used in the equations of motion
CYtotal ¼ CYb ðM; aÞb þ Dr CY ðM; dr ; aÞ
Symbol Definition
þ b=ð2VÞ½CYp ðM; aÞPs þ CYr ðM; aÞRs
V ¼ ðU; V; W Þ Translation velocity in the body-fixed frame
x ¼ ðP; Q; RÞ Angular velocity (roll, Pitch, Yaw)
f; y; c Euler angles (roll, Pitch, Yaw) Cltotal ¼ Clb ðM; aÞb þ Da Cl ðM; da ; aÞ
Fx ; Fy ; Fz X ; Y ; and Z components of aerodynamics and þ b=ð2VÞ½Clp ðM; aÞPs þ Clr ðM; aÞRs
thrust forces  ÞC total ðc=bÞcos a
 ðXcg  Xcg Y
L; M; N Rolling, Pitching, and Yawing moments
pN ; pE ; h Position components in the Earth-surface fixed
frame total
CM ¼ CM ðM; df ; aÞ þ De CM ðM; de ; aÞ
(North, East, Altitude)
þ c=ð2V ÞCMq ðM; aÞq
m Mass of the aircraft
 ÞðC total cos a þ C total sin aÞ
þ ðXcg  Xcg
g0 Gravity constant L D
ci ; i ¼ 1; 2; y; 9 Mass properties of the aircraft (see Stevens &
Lewis, 1992)
CNtotal ¼ CNb ðM; aÞb þ Dr CN ðM; dr ; aÞ
þ b=ð2VÞ½CNp ðM; aÞPs þ CNr ðM; aÞRs
 ðX  X  ÞC total ðc=bÞcos a
cg cg Y
side force are most conveniently expressed in the
stability-axis system and are transformed into the
body-fixed axis to be used in Eqs. (3) and (5). Those
aerodynamic forces and moments are determined from deflection df ; and angle-of-attack a: Fig. 2 shows the
the nondimensional aerodynamic static and dynamic variation of lift coefficient CL ðM; df ; aÞ as a function of
stability derivatives, which are often obtained from the df and a at two different Mach numbers. Multi-
extensive theoretical and experimental works and given dimensional linear interpolation is used to obtain the
in the format of a numerical look-up table. lift coefficient at an arbitrary flight condition.
The basic flight vehicle frame used in this paper is a
conventional fixed-wing jet aircraft that Korea Aero- 2.3. Control surface servo and engine modeling
space Research Institute has designed for internal
research and development purposes and its physical First-order system is adopted to model the control
properties are summarized in Table 2. The number of surface servo. That is, the transfer function between the
static and dynamic stability derivatives including those commanded deflection dc and the actual deflection d is
for the engine is 22 and most of them are a function of a 1
Mach number, angle-of-attack, and control surface d¼ dc ðsÞ; ð7Þ
tc s þ 1
deflection angle. The relation between the stability
derivatives and force and moment coefficients for this where tc is the time constant, and the actual deflection is
particular aircraft model is summarized in Table 3. For limited within 720 .
example, CL ðM; df ; aÞ denotes the lift coefficient of the The relation between the commanded throttle open-
basic airframe at a particular Mach number M; flap ing dt and the actual opening dt is modeled similarly as
 ARTICLE IN PRESS
604 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

Lift Coefficentat Mach Num. = 0.1 (CL) Lift Coefficent at Mach Num. = 0.2 (CL)

2 3

1.5
2
1

CL
CL

0.5 1

0
0
-0.5
-1 -1
30 30
20 20 20 20
10 10
10 10
0 0
Flap angle 0 -10 Flap angle 0 -10
ALPHA ALPHA

Fig. 2. Lift coefficient at various flight conditions.

THRUST AT MACH = 0.1 THRUST AT MACH = 0.4

10000 12000

8000 10000
8000
THRUST

THRUST

6000
6000
4000
4000
2000 2000
0 0
2 2
1.5 1 1.5 1

1 0.5 1 0.5

EPR 0.5 0 APR EPR 0.5 0 APR

Fig. 3. Engine thrust at various flight conditions.

above. The actual throttle opening dt determines the 3. Control system design
engine power level (EPL) and this process is assumed to
be a first-order. Hence, the whole process can be 3.1. Reference flight condition and linear equations of
modeled as motion

dt EPL 1 1 Following Stevens and Lewis (1992), classical control

 ðsÞ ðsÞ ¼ ; ð8Þ design method is used to obtain various components of
dt dt tt s þ 1 tEPL s þ 1
the stability and control augmentation systems. First,
the equations of motion with all or some of the
where tt and tEPL ; respectively, denote the time derivative terms set to zero have to be solved in
constants of the throttle and engine. conjunction with the constraint equations that define a
For the static engine map that relates EPL to the particular reference flight condition. Thus obtained
thrust force T acting on the aircraft, two sets of lookup solutions are used as a part of initial conditions
tables are used. One is used to determine the engine for the nonlinear simulation. For example, Fig. 4 shows
pressure ratio (EPR) as a function of EPL, Mach the trim elevator deflection angle and EPL that are
number, and altitude h; and the other to determine the required for the steady state, descending flight with the
engine thrust T as a function of EPR, ambient pressure flight path angle of g0 ¼ 2:5 at various speed and
ratio (APR), and Mach number Fig. 3 shows the altitudes.
variation of the thrust as a function of EPR and APR at For obtaining a set of linear equations of motion, the
two different Mach numbers. nonlinear equations of motion are numerically per-
 ARTICLE IN PRESS
T.S. No et al. / Control Engineering Practice 13 (2005) 601–612 605

Elevator deflection Engine Power Level

-5 9
elevator deflection (deg.)

-10 8

-15 7

EPL
-20 6

-25 5

-30 4
1500 1500
1000 270 1000 270
260 260
500 250 500 250
240 240
Altitude (ft) 0 230 Velocity (ft/sec.) Altitude (ft) 0 230 Velocity (ft/sec.)

Fig. 4. Elevator angle and engine power level at various trim flight conditions.

turbed around the reference flight condition. In this Lateral mode (yaw, roll, y-direction)
paper, the steady state, straight, and level flight at the 2 3
0:1953 0:1286 0 0:0643 0:9979
speed of 250 ft/s is chosen as a reference flight condition. 6 7
6 0 0 0 1 0:0228 7
This choice is only for designing the control systems. 6 7
Trimmed flight much below this speed would be possible x’ lat ¼ 6
6 0 0 0 0 1 7xlat
7
6 7
if the elevator deflection is not limited and the engine 4 4:7633 0 0 3:1885 0:8535 5
power is increased significantly. 2:1426 0 0 0:2892 0:6621
For the aircraft model used in this paper, it is found 2 3
0 0:0006
that the motion is separated into longitudinal and lateral 6 7
modes (which is very typical in the conventional type 6 0 0 7
6 7
aircraft). These are 6
þ6 0 0 7ulat ; ð10Þ
7
Longitudinal mode (pitch, x- and z-direction, engine 6 7
4 0:0137 0:0069 5
dynamics)
0:0009 0:1031
2 3
0:0239 20:643 32:193 0 0 0 2 3 2 3
6 0:0010 1:0856 0:0056 0:9215 0 0 7 b side slip
6 7 6 7 6 7
6 7
6 0 0 0 1 0 0 7 6 f 7 6 roll angle 7
x’ long ¼ 6
6 0:0006
7xlong 6 7 6 7
0 7 where; xlat ¼ 6 7 6 7
6
6
5:9650 0:0034 2:3991 0 7
7 6 c 7 ¼ 6 yaw angle 7;
4 0:0436 249:762 249:762 0 0 0 5 6 7 6 7
4 p 5 4 roll rate 5
0 0 0 0 0 2
2 3 r yaw rate
0:0813 0:0218
6 0 0:0012 7
da aileron deflection
6
6
7
7 ulat ¼ ¼
6 0 0 7 dr rudder deflection
þ6
6 0
7ulong ; ð9Þ
6 0:0374 7
7
6 7
4 0 0 5 3.2. Controller design and response
2 0
Controllers for pitch and roll attitude stabilization,
2 3 2 3 attitude orientation, and sideslip suppressor for coordi-
v speed
6 a 7 6 angle-of-attack 7
7 6 nate turns are designed and integrated with the basic
6 7
6 7 6 7 airframe as shown in Fig. 5. The controllers for speed,
6 y 7 6 pitch angle 7
where; xlong ¼6 7¼6 7; flight path angle, and heading orientation constitute the
6 7 6
q 7 6 pitch rate 7
6 7 outer loops. Proportional-integral-derivative controllers
6 7 6 7
4 h 5 4 altitude 5 and lead/lag compensator are effectively combined to
EPL engine power level assure enough gain and phase margin. Fig. 6 shows the
Bode frequency and magnitude response of the flight
dt throttle opening
ulong ¼ ¼ path angle and heading orientation control loops,
de elevator deflection respectively.
 ARTICLE IN PRESS
606 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

Trim Value limit Q

VT +
- 20 s + 4 10
s +10
+ θ
Vc + s
Speed Controller Q 100s + 60
Engine VT
s
θ Pitch SAS Trim Value limit
γ
57.3
+
− 20
γ - 20( s + 0.2)(s + 1.4) +
- 40( s + 0.6)( s + 0.1) s + 20
57.3 + s ( s + 14)
γc + s 2 ( s + 10)
Elevator
Flight Path Controller
Pitch Attitude Controller ψ
Trim Value limit
ψc + 40 ( s + 0.1)( s + 1.5) φ
+ +
- s ( s + 15) 200 + − 20
1.0 - +
s + 20
Heading Controller φ -
P
Roll 50 Aileron
Attitude P
β
Controller Roll SAS Trim Value limit
+
150 + − 20 R
β s + 20
+
-
s
10 Rudder 6DOF
s +1 Nonliner
Aircraft
Yaw SAS
R Model

Fig. 5. Stability augmentation, speed and flight direction control system.

Fully 6-DOF nonlinear simulation is performed to where ðex ; ey ; ez Þ are the unit direction vectors of the
investigate the response of the aircraft including all the fixed frame. If both the missile and the target maintain
control loops to a command input. Nonlinear equations their respective speed and flight direction, the distance
of motion, aerodynamic force and moment, and engine vector between them at yet unspecified future time tf can
models are used as discussed in Section 2. Fig. 7 shows a be written as:
typical response of three control channels (speed, flight d ¼ ðd   dÞ þ ðv  vÞt
tgo go
path angle, and heading) to a step input.
¼ Mx ex þ M y ey þ Mz ez ; ð13Þ
where tgo is the time-to-go until the future time tf : Time-
4. Guidance law for trajectory-tracking to-go tgo can be written as
tgo ¼ tf  t: ð14Þ
4.1. Zero effort miss
This vector is often referred to as the zero effort miss
Let us place an imaginary target on the desired vector and used in obtaining the various types of
reference trajectory and assume that it flies along the proportional guidance laws (Zarchan, 1992).
trajectory ideally in terms of its position and velocity.
Then the aircraft is treated as a missile that tries to 4.2. Derivation of guidance law
intercept the imaginary target. Use (d,v) and (d,v) to
denote the position and velocity vectors of the aircraft If the chasing aircraft keeps d tgo small for sufficiently
and the target, respectively, short tgo ; it implies that the aircraft is following the
reference trajectory with a small error. A scalar,
d ¼ dx e x þ dy e y þ dz e z ; v ¼ v x e x þ v y e y þ v z e z ; ð11Þ
positive-definite function is introduced as follows:
d  ¼ dx ex þ dy ey þ dz ez ; v ¼ vx ex þ vy ey þ vz ez ; X ¼ 12 d tgo dd tgo
ð12Þ ¼ 12ðMx2 þ My2 þ Mz2 Þ: ð15Þ
 ARTICLE IN PRESS
T.S. No et al. / Control Engineering Practice 13 (2005) 601–612 607

Gm = 7.6 dB (at 5.74 rad/sec) , Pm = 60.6 deg (at 1.18 rad/sec) 15

∆ VT (ft/sec)
150
10
100
Magnitude (dB)

50
5

0 0
0 5 10 15 20 25 30
-50 Time (sec)
(a) Speed
-100
-90
1.5

∆ ψ (deg.)
-180
Phase (deg)

1
-270
0.5
-360
0
0 5 10 15 20 25 30
-450 Time (sec)
10-4 10 -3 10 -2 10 -1 10 0 10 1 10 2
(b) Heading
(a) Frequency (rad/sec)
1.5
Gm = 15.6 dB (at 3.13rad/sec), Pm = 71.3 deg (at 0.49 rad/sec)

∆ γ (deg.)
1
100
0.5
Magnitude (dB)

50 0

0
-0.5
0 5 10 15 20 25 30

-50
Time (sec)
(c) Flight path
-100
0 Fig. 7. Response of three control channels to step input, (a) Speed; (b)
Heading; (c) Flight path.
-180
Phase (deg)

-360

-540 where c and b denote, respectively, yaw and sideslip

angles.
-720
-3 -2 -1 0 1 2
Replacing s with c in Eq. (17) and substituting it into
10 10 10 10 10 10
(b) Frequency (rad/sec)
Eq. (16), one gets

Fig. 6. Bode frequency response of control loops, (a) Flight path angle dX
control loop; (b) Heading orientation control loop. ¼ Mx ð’vx  v’ cos c cos g
dt
þ vc ’ sin c cos g þ v’g cos c sin gÞtgo
Differentiation of Eq. (15) yields þ My ð’vy  v’ sin c cos g
dX ’ cos c cos g þ v’g sin c sin gÞtgo
¼ Mx ð’vx  v’x Þtgo þ My ð’vy  v’y Þtgo  vc
dt
þ M ð’v þ v’ sin g þ v’g cos gÞt : ð19Þ
þ Mz ð’vz  v’z Þtgo :
z z go
ð16Þ

The velocity vector v of the aircraft in the fixed frame A control frame ðev ; ec ; eg Þ is introduced where ev
is denotes a unit direction vector along the current velocity
vector, eg is a unit direction vector perpendicular to ev
v ¼ vx ex þ vy ey þ vz ez and positive in the sense of increasing the longitudinal
¼ v cos g cos sex þ v cos g sin sey  v sin gez ; ð17Þ flight path angle, and ec completes the right-handed
system and denotes a unit vector along the direction of
where v is the aircraft speed, g is the longitudinal flight increasing the yaw angle c: Then, ðMx ; My ; Mz Þ; the
path angle, and s is the lateral flight path angle. If it is components of the zero effort miss vector in the fixed
assumed that the aircraft performs a perfect coordinate frame, can be redefined in the control frame as follows:
turn, then the lateral flight path angle, s can be
approximated to Mv ¼ Mx cos c cos g þ My sin c cos g  Mz sin g;
s ¼c þ b Mc ¼  Mx sin c þ My cos c;
E c; ð18Þ Mg ¼ Mx cos c sin g þ My sin c sin g þ Mz cos g: ð20Þ
 ARTICLE IN PRESS
608 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

Similarly, ð’vv ; v’c ; v’g Þ denote the target acceleration natural choice appears to be
vector in the control frame. Then Eq. (19) can be N
simplified to vc ¼ v þ tv Mv þ tv v’v ; ð28Þ
tgo
dX
¼ ð’vv  v’ÞMv tgo þ ð’vc  vc’ cos gÞMc tgo N tc tc
dt cc ¼ c þ Mc þ v’ ; ð29Þ
tgo v cos g v cos g c
þ ð’vg þ v’gÞMg tgo : ð21Þ
N tg tg
In order to use the Lyapunov stability theorem gc ¼ g  Mg  v’g : ð30Þ
tgo v v
(Slotine & Li, 1991), negative definiteness of Eq. (21)
must be assured. One way to achieve this, among many One may easily note that the proposed form of the
alternatives, is to set guidance laws feature several important aspects of the
trajectory-tracking problem. Firstly, the commands are
dX
¼ 2NX; ð22Þ updated in feedback manner. The current flight speed
dt and directions are corrected proportional to the zero
where N is a positive constant. Next step is to introduce effort miss distance vector and target acceleration vector
the systems dynamics into Eq. (21) so that it will satisfy in the control frame. This is a very similar scheme as in
Eq. (22). In this case, the dynamics of the three control the conventional proportional guidance laws (Zarchan,
channels for speed v; flight path angle g; and heading 1992). Secondly, the system characteristics, as repre-
angle c must be identified. Since the explicit and sented by time constantsðtv ; tc ; tg Þ; are appropriately
analytical expression of such dynamics is practically reflected in the guidance commands. The magnitude of
impossible due to the nonlinearity and complexity of the commands is modulated according the speed of the
complete system, it is assumed that they can be reduced system dynamics.
or approximated to a simple lower-order system. From Lastly, one would note that the guidance parameters
the step response shown in Fig. 7, it is assumed that the N=tgo act like a proportional gain. A relatively large N
speed and flight path angle responses with the attitude needs to be used if one wants the flight vehicle to follow
stabilization loops closed can be approximated to those the reference trajectory quickly. A relatively small tgo
of the first-order systems as follows: should be employed if the precision is a more important
Speed control loop: measure of tracking performance. While the use of any
positive N and tgo is not restricted for application to a
1
v’ ¼ ðvc  vÞ ð23Þ linear system (No et al., 2001a, b), these parameters may
tv be selected via some numerical experiments in con-
Longitudinal flight path angle control loop: sideration of the system characteristics such as actuator
1 limit and controllers’ bandwidth for application to a
g’ ¼ ðgc  gÞ ð24Þ nonlinear system. For example, too large N (or too
tg
small tgo ) will lead to the actuator saturation due the
Heading angle control loop: unrealistically large input commands.
’ ¼ 1 ðcc  cÞ;
c ð25Þ
tc
where ðtv ; tc ; tg Þ are the time constants of each control 5. Implementation and 6-DOF nonlinear simulation
loop and ðvc ; cc ; gc Þ are the guidance command input for examples
corresponding control loops.
Note that the Lyapunov function X can be equiva- 5.1. Implementation
lently rewritten as
Complete system for the arbitrary trajectory-tracking
X ¼ 12ðMv2 þ Mc2 þ Mg2 Þ ð26Þ control may be configured as shown in Fig. 8. Stability
and substitute Eqs. (23)–(25) into Eq. (22) to get
dX
¼ ð’vv  v’ÞMv tgo þ ð’vc  vc’ cos gÞMc tgo
dt
þ ð’vg þ v’gÞMg tgo
¼  2NX
¼  NMv2  NMc2  NMg2 : ð27Þ
A set of guidance commands ðvc ; cc ; gc Þ is selected so
that it satisfies Eq. (27). Among several alternatives, the Fig. 8. Trajectory-tracking control system.
 ARTICLE IN PRESS
T.S. No et al. / Control Engineering Practice 13 (2005) 601–612 609

augmentation and flight path controllers shown in Fig. 5

form the inner loops and the guidance loop, that is, the
command generator forms the outer loop. Given a 2000
reference trajectory to follow, the guidance commands Aircraft Flight Trajectory
1500
are computed according to the procedures discussed in

Altitude (ft)
Section 4, specifically Eqs. (28)–(30). One should note
1000
that the position and velocity errors are effectively used Reference Trajectory
to compute the speed and flight direction commands 500
that are readily acceptable to the flight control systems
of many existing aircrafts. 0
-20
0 0
5.2. Simulation examples 1
20 2
3 × 104
East Distance (ft) 4
40 5 Downrange (ft)
5.2.1. ILS landing
Typical landing procedure of a commercial aircraft Fig. 10. Flight trajectory during ILS landing.
consists of two phases, the glide slope capture and the
flare modes (Shue & Agarwal, 1999). Fig. 9 shows the
geometry for the landing. During the glide capture 8 35
mode, the aircraft descends along the predetermined 6 30
glide path. Conventionally, the flight path angle g is

Lateral Distance Error (ft)

4 25
indirectly controlled by reducing the glide path devia- Altitude Error (ft)
tion error d: Note that d is related to g as below: 2
20

d’ ¼ VA sinðg  gR Þ; ð31Þ 15 Altitude Error

0
10
where VA is the aircraft speed and gR is the reference -2
flight path angle. 5
After passing the decision height, the aircraft reduces -4
0
Lateral Distance Error
its height exponentially to assure a smooth touch down -6 -5
with a small sink rate. During the flare mode, the height
is directly controlled so that the vehicle follows the -8 -10
0 20 40 60 80 100 120 140 160
reference height profile hðtÞ represented by Time (sec)
t
hðtÞ ¼ h0 et ; ð32Þ Fig. 11. Position error during ILS landing.

where h0 denotes a decision height and the time constant

t is determined considering various conditions such as and velocity, two modes can be easily combined into
the distance between the runway threshold and the one.
touchdown point (Stevens & Lewis, 1992). In the simulation example, the aircraft is flying
Therefore, the conventional ILS requires two separate horizontally and entering the descending mode with
controllers to be designed, one for the glide slope the initial speed of 270 ft/s and the heading error of 3 ,
capture mode, and the other for the flare mode. If one and then it gradually reduces its speed to 250 ft/s. At the
notices that the desired speed schedule, reference flight decision height of 45 ft, it reduces its height exponen-
path angle, and the reference height profile are sufficient tially while keeping the speed constant.
to generate the reference trajectory in terms of position Fig. 10 illustrates the three-dimensional flight trajec-
tory during approach and landing. After the initial
transition, the aircraft follows the desired path. Time
VA histories of the position errors from the glide path are
(γ − γ R ) shown in Figs. 11 and 12, respectively. There is a short
d transient period due to the change in the reference
Glide slope
trajectory from the straight one to the exponential one.
t = t0
Although time histories of other relevant variables such
Flare path h0 as sink-rate, pitch, angle-of-attack, etc. are not shown
t = t f −γ −γ R here, they are well under control.
Touchdown Glide slope
transmitter
D 5.2.2. Reconnaissance and surveillance maneuver
This particular maneuver is taken from Boyle and
Fig. 9. ILS landing and flare geometry. Chamitof (1999) where the quasi-trimmed flight condi-
 ARTICLE IN PRESS
610 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

Velocity (ft/sec)
2 2 300

X-LVLH
150
1.5
1 0
1
-150
0.5 Yaw Angle Error
-300
0 0 50 100 150 200 250 300 350 400 450 500
ψerror (deg.)

0
γerror (deg.)

Velocity (ft/sec) Velocity (ft/sec)

Flight Path Angle Error 300

Y-LVLH
-1 -0.5 150
-1 0
-2 -150
-1.5
-300
-2 0 50 100 150 200 250 300 350 400 450 500
-3 20
-2.5

Z-LVLH
10
-4 -3
0 20 40 60 80 100 120 140 160 0
Time (sec)
-10
0 50 100 150 200 250 300 350 400 450 500
Fig. 12. Flight direction error during ILS landing.
Time (sec)

Fig. 14. Velocity time histories during reconnaissance and surveillance

maneuver.

3000
histories of the velocity components in the fixed frame
Altitude (ft)

2500
2000 are shown in Fig. 14, from which one would easily
1500
identify the maneuver sequence. That is, accelerated
ascent, turn with the constant turning and ascent rate,
1000
2 straight level flight, turn with the constant turning and
descent rate, and then accelerated ascent.
1
4
× 10 0 4
3
5.2.3. Helical trajectory-tracking
-1 2 4
× 10 The example trajectory taken from Kaminer et al.
1
-2
0 (1998) is a helix on which the velocity of the ideal vehicle
East Distance (ft) -1 North Distance (ft)
in the fixed frame can be specified as follows:
Fig. 13. Flight trajectory during reconnaissance and surveillance
v ¼ vc cos gc cosðc’ c tÞex þ vc cos gc sinðc
’ c tÞey
maneuver.

 vc sin gc ez ; ð33Þ

tion is used to find a set of maneuver commands, and ’ c denote desired speed, flight-path
where vc ; gc ; and c
then the robust tracking controller is designed to angle, and yaw rate, respectively, on the reference
account for model uncertainty and disturbance rejec- trimming trajectory. Eq. (33) is used in the reference
tion. The example maneuver sequence used in this trajectory block in Fig. 8 and the guidance commands in
present paper may be summarized as follows: (1) wings- speed, flight path angle, and heading angle are
level accelerated ascent at 2 ft/s2; (2) a 360 ascending calculated by using Eqs. (28)–(30).
left turn with a constant ascent rate of 5 ft/s and a turn In the numerical simulation, vc ¼ 270 ft=s; gc ¼ 2 ;
rate of 2 /s; (3) a straight and level segment of 80 s ’ c ¼ 2 =s are used and the actual vehicle is initially
c
duration; (4) a 360 descending right turn with a flying straight with constant speed of 270 ft/s. To
constant descent rate of 5 ft/s and a turn rate of 2 /s demonstrate the effectiveness of the proposed method,
and (5) a wings-level accelerated descent at 2 ft/s2. The two different approaches are compared. In the first
aircraft enters the maneuver from the steady state, approach, the reference speed and flight angles are
straight flight at the speed of 270 ft/s and the altitude of directly applied to the speed and flight direction
1500 ft. controllers shown in Fig. 4. In the second method, the
Fig. 13 shows the flight trajectory during maneuver. guidance commands are calculated using Eqs. (28)–(30)
This maneuver is qualitatively the same as that of Boyle and used in the trajectory-tracking control system in
and Chamitof (1999) except that the specific numbers Fig. 8. Fig. 15 shows the flight trajectory. As can be seen
are different because the basic airframes used in each of from Fig. 16, the position tracking performance of the
the examples are different. The difference in the desired new approach is superior to that of direct application of
and the achieved trajectories is rarely discernible. Time speed and flight path direction as input commands. Fig.
 ARTICLE IN PRESS
T.S. No et al. / Control Engineering Practice 13 (2005) 601–612 611

3-D Aircraft Trajectory pitch and roll rates reach their steady-state values of
pss ¼ 0:21 =s and qss ¼ 0:57 =s: One should note that
the pitch angle increases as the vehicle ascends. This is
7000 due to the fact that the vehicle should pitch up to
6000 increase its angle-of-attack since the atmospheric density
decreases as the vehicle goes higher.
Altitude (ft)

5000
4000
3000
2000 6. Conclusions and remarks
1000
2
In this paper, a rather complete description of 6-DOF
1 1
4 0.5 nonlinear flight simulation has been presented. Fully
× 10 0 0 4
-0.5 × 10 nonlinear equations of motion and multi-dimensional
East Distance (ft) -1 -1 numerical look-up tables for the aerodynamic force
North Distance (ft)
and moment were used. A set of linear equations of
Fig. 15. Helical flight trajectory. motion was obtained by using numerical perturbation
method. Classical root-locus and Bode frequency
500 40
method were used to design the attitude stabilization,
Lyapunov speed and flight path controllers. The concept of
450 Direct
35 zero effort miss, which is frequently used in the
400
Lyapunov distance error (ft)

30
problems of a missile–target intercept, was effectively
Direct distance error (ft)

350 used to propose a guidance law that is applicable to

300 25 the problem of arbitrary trajectory-tracking control. In
250
the proposed guidance scheme, the commands are
20
given in terms of speed, flight path and heading
200
15 angles so that they may be easily fitted with the existing
150 control configuration. However, they fully reflect
10
100 any positional and velocity tracking errors, which is
50 5 essential for the precise match between the desired
and the achieved trajectories. Results from the non-
0 0
0 100 200 300 400 500 600 linear simulation of the ILS automatic landing, recon-
Time (sec)
naissance and surveillance maneuver, and helical trajec-
Fig. 16. Time history of helical trajectory-tracking error. tory-tracking maneuver demonstrate the applicability
of the new method. In those three examples, controller
10 35
architecture was not altered at all or controllers
p : Roll Rate
gains were not fine-tuned. Only the guidance para-
8 30 q : Pitch Rate meters N and tgo for the outer loop were varied. This is
6 possible since at least some portion of the system
Roll and Pitch Rates (deg./sec)

Bank and Pitch Angles (deg.)

25
4 (controller) information is used in the design process of
2 20 guidance laws.
φ
0 15
-2 10 q p
-4 Acknowledgements
5 θ
-6
0 φ : Bank Angle Part of this work was performed while the first author
-8 θ : Pitch Angle was at the University of Sydney as a visiting scholar, and
-10 -5 the support from the Korea Research Foundation
0 100 200 300 400 500 600
Time (sec) (KRF-2001-013-E00028) is greatly acknowledged.
Fig. 17. Time histories of attitude angles and rates during helical
trajectory tracking.
References
17 shows the time histories of attitude angles and rates. Al-Hiddabi, S. A., & McClamroch, N. H. (2002). Aggressive
The vehicle maintains a constant bank angle of longitudinal aircraft trajectory tracking using nonlinear control.
approximately fss ¼ 16:5 for its coordinate turn, and Journal of Guidance, Control, and Dynamics, 25(1), 26–32.
 ARTICLE IN PRESS
612 T.S. No et al. / Control Engineering Practice 13 (2005) 601–612

Boyle, D. P., & Chamitof, G. E. (1999). Autonomous maneuver No, T. S., Chong, K. T., & Rho, D. W. (2001a). A lyapunov fun-
tracking for self-piloted vehicles. Journal of Guidance, Control, and ction approach to longitudinal control of vehicles in a pla-
Dynamics, 22(1), 58–67. toon. IEEE Transaction on Vehicular Technology, 50(1),
Fujimori, A., Kurozumi, M., Nikiforuk, P. N., & Gupta, M. M. 116–124.
(2000). Flight control design of an automatic landing flight No, T. S., Cochran Jr., J. E., & Kim, E.-G. (2001b). Bank-to-turn
experiment vehicle. Journal of Guidance, Control, and Dynamics, guidance law using lyapunov function and nonzero effort miss.
23(2), 373–376. Journal of Guidance, Control, and Dynamics, 24(2), 255–260.
Gibbens, P.W. (1998). Maneuver Controller Design for an F-111C Schultz, R. L. (1990). Three-dimensional trajectory optimization for
Flight Dynamics Model. Air Operations Division, Aeronautical and aircraft. Journal of Guidance, Control, and Dynamics, 13(6),
Maritime Research Laboratory, DSTO-RR-0129. 936–943.
Heymann, V. I., & Ben-Asher, J. Z. (1997). Aircraft trajectory Shue, S.-P., & Agarwal, R. K. (1999). Design of automatic landing
optimization in the horizontal plane. Journal of Guidance, Control, systems using mixed H2/HN control. Journal of Guidance, Control,
and Dynamics, 20(6), 1271–1274. and Dynamics, 22(1), 103–114.
Jung, Y. C., & Hess, R. A. (1991). Precise flight-path control using a Slotine, J.-J. E., & Li, W. (1991). Applied nonlinear control. Upper
predictive algorithm. Journal of Guidance, Control, and Dynamics, Saddle River, NJ, USA: Prentice-Hall.
14(5), 936–942. Stevens, B. L., & Lewis, F. L. (1992). Aircraft control and simulation.
Kaminer, I., Pascoal, A., Hallberg, E., & Silvestre, C. (1998). USA: Wiley.
Trajectory tracking for autonomous vehicles: An integrated Wu, S.-F., Engelen, C. J. H., Babuska, R., Chu, Q.-P., & Mulder, J. A.
approach to guidance and control. Journal of Guidance, Control, (2003). Fuzzy logic based fill-envelope autonomous flight control
and Dynamics, 21(1), 29–38. for an atmospheric re-entry spacecraft. Control Engineering
Marconi, L., Isidori, A., & Serrani, A. (2002). Autonomous vertical Practice, 11, 11–25.
landing on an oscillating platform: An internal-model based Zarchan, P. (1992). Tactical and strategic missile guidance (2nd ed.).
approach. Automatica, 38(1), 21–32. Washington, DC, USA: AIAA.