by
A thesis submitted to
the Faculty o f Graduate Studies and Research
in partial fulfilment of
the requirement for the degm of
Master of Engineering, Aerospace
Carleton University
Ottawa, Ontario
January 24,1999
The author bas granted a non L'auteur a accord une licence non
exclusive licence dowing the exclusive pemettant la
National Libmy of Canada to Bibliothque nationale du Canada de
reproduce, loan, distribute or sell reproduire, prter, distribuer ou
copies of this thesis in rnicrofonn, vendre des copies de cette thse sous
paper or electronic formats. la forme de rnicrofiche/nlm, de
reproduction sur papier ou sur format
lectronique.
Figures Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . "
..
List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vil
...
Aeronauticai Variable Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . wil
Chapterl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
DynamicsofAircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 htroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Chapter2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Aircraft Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1 Detemination o f Dominant Stability Coefficients . . . . . . . . . . . . . 9
2.2 Longitudinal Stability Denvatives . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Longitudinal Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Longitudinal Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Stability Derivative of Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Laterai Stability Denvatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Laterd Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Lateral Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Chapter4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Mathematical Mode1 of 4ircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 Mathematical Representation of the Dynamics of Aircraft . . . . . . 24
4.2 The Aerodynamic Forces and Moments acting on Aircraft . . . . . . 24
Chapter5 ......................................................35
LQR Control of Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.1 Solution of the Linear Quadratic Regulator (LQR) . . . . . . . . . . . . 35
Chapter6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Fuuy Logic Control of Aircrafl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
61 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.2 Fu Logic Control Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.2.1 Fuuy Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
62 . 2 Operation on Fuzzy Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.2.3 CnspSet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2.4 Fuzzfication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2.5 Degree of Fulfillment . . . . . . . . . . . . . . . . . . . . . . . . .. . . . .  48
6.2.6 Implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2.7 Aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2.8 Defuzzification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2.9 Fu Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.2.10 Membership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Chapter7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Simulation and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.1 Longitudinal Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Laterd Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Chapter8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Appendix:A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
A .1 Characteristics of Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
A.2 Aerodynamics and Control Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Appendix:B .................................................... 93
F u v y Logic Programmuig For Aircraft Motion . . . . . . . . . . . . . . 93
Nurnber Figures Definition Page
Euler angular rotation in roll time in dynamic Iateral motion wth LQR and
fuay logic controller.. ....................................................................................... 79
Roll angular rates versus time in dynarnic lateral motion with the LQR and ft
Iogic controller.................................................................................................. 80
Yaw angular velocities versus t h e in dynarnic lateral motion with the LQR and
fuzzy lo@c controller........................................... ............................................. 31
Angular deflection of mdder versus time in dpamic lateral motion with the LQR
and b z y logic controller.................................................................................. 82
.Angular deflection of aileron versus time in dpamic lateral motion with the LQR
and f k z y logic control...................................................................................... .S
Number Tables
Error versus cnsp logic values....................................................................... 47
Error versus fiizzy logic values...................................................................... 50
Error versus fuuy logic control action values................................................ 50
Rule base for throttle membership function setting changes............................ 52
Logic control values for throttle setting membership functions...................... 52
Rule base for pitch angle changes................................................................... 53
Logic control values for pitch angle variation of membership functions.......... 53
Rule bases for roi1 angle and rudder commands.............................................. 54
Characterstics of hypothetical aircraft.......................................................... 88
. .
Aerodynarnic and cr~stroldenvatives............................................................. 90
Symbol List of Acronyms
rU hificial intelligence
CG Center of gavity
dot Refers to the rate of change with respect to time
FLC FU logic controI
YB Neoative big
'CM Negative medium
!?S Negative srnail
PS Positive smali
PM Positive medium
PB Positive big
T Refers to a value at trim condition
zero Indicates a vaIue at zero time
ky.z) Illustrates the body axes (mm and cm)
Sci / d a , i = L,M,N
Dynamic derivative with respect to arate of change parameter.. ..........()
C.
'Sa Static derivative with respect to aileron deflection................................... (1
Zci/3a,i = L, N,y
Cci / d G e , i = M,x,z
c  Static derivative with respect to rudder deflection................................... (4
'6r
d c i / d 6 .,i = L,N,y
Density of air.................................................................................(Wm3)
Aileron effectiveness............................................................................... (1
Rudder effectiveness.............................................................................. (4
Venical flight path angle with respect to horizontal plane.. ........(rad or deg)
Horizontal tlight path angle with respect to vertical plane.. .......(rad or deg)
Subscrkts
Superscript refers to rate of change with respect to time
Subscript O refers to zero time value
Chapter 2
Aircraft Modeling
linearly dependent on a and & . The mathematical modeiing of the aircraft and the
consideration of al1 parameters in longitudinallateral motion makes it possible to solve the
problem of aerodynamic forces and moments. Primary consideration is given to the non
lineanty motion and it's dependence of time. As these two motions are strongly coupted, it
was necessary to analyze six degree of fieedom motion.
To simpliw the analysis of motion of the aircraft, the following equations are
considered:
the external forces and moments
the auxiliary equations
. the physical charactenstics
the stability and control derivatives and the initial conditions Cg].
. The forces and moments acting on the aircrafi are determined in terms of
dimensionless aerodynamic coefficients as:

side force, Y = q Sc,.
The dimensionless force coefficients have been specified for wind axes; the bodyaxes
forces coefficients c,, c , , and c, are also in cornmon use. The moments coefficients rnay
be specified in wind, stability, or body axes, but the same symbols will be used in each
case. The various dirnensionless coefficients c, , c , . ..., c, are the dependent on the
aerodynamic angles alpha and beta and less dependent on other variables. The dependence
on the rates of changes of the aerodynamic angles and on the components p, q and r of the
center ofgravity of the aircraft angular velocity is elaborated in equations (26) and (45).
The coefficients are dso dependent on control surface deflections and other factors such
as engine power level. configuration effects ( e.g; landing gear, extemal tanks. etc)
otherwise, the aircraft wouldn't be controlIable GO, 48. 491.
The rnost important longitudinal stability derivatives are
where:
CL. , c,.  darnping derivatives that descnbe the incrernent in lift due to the
a 4
alpha dot (dal dt) and pitch rate. The alphadotrrefers to the rate of
change of angle of attack with respect to time.
c ,,q , .,c  damping derivatives in moments equations that descnbe the increment
a
in lie due to the alphadot and pitch rate. The cornputer analysis study of air vehicle
motion in straight and level flight depends on the variables which determine the degree of
stability. The dynamic derivatives include the angular rotations (p, q, r), darnping
1 1
due to the time rate of change of aerodynamic angles ( a. ) . Some control derivatives
pitch, alpha, rate of change: and c,,, ,and c=, , static derivatives which correspond to
Q> would be presented; but if it is for c , , ~and c,. ,then the rudder setting must be
P
disturbed and the aircraft response is recorded. The sign for aerodynamic control surface
perturbation c m be either negative or positive. The effect of dynaniic derivatives with
respect to yaw and beta rate parameters can be analyzed on the longitudinal aircraft
motion by introducing lateral perturbation of the rudder angle .
The result indicates that both longitudinal and lateral perturbations have an effect on the
aircraf? motion; the most significant is the overall effect o f perturbing the rudder for the
straight and level flight, as the pemirbation could make the air vehicle drift away ffom the
original flight path. The c,, has more effect on alpha than cMrand c,. which indicate
B
higher degrees of significance for static derivatives. The rate of roll (p) and the rate of
change of sideslip p are high cornpared to the rate of yaw (r). The rate of pitch (q) is
12
relatively small. The denvatives of various aerodynamic reactions due to angular velocities
(p. q, r) represent the quasisteady effects. These derivatives forrn part of the dynamic
results obtained during standard windtunnel experiments of oscillations around a fixed
axis. where the output is always an expression composed of more than derivatives such as
.
fiom O to 40' , while Figure 5.2 shows the lowsubsonic cross derivatives
( c ,p f c  c,. cosa ) for a swept wing fighter aircraft [2]. Nonlinear
s i n a ).and (cLr
P B
variation of some of the cross coupling denvatives is shown in Figure 5.3. Nonlinear
behavior of the dynamic crosscoupling moment derivatives due to pitching ( csq+ .c, )
u
even low angle of attack in case of yawing, where sideslip angle is nonzero [LOI.
The basic test to determine directional stability and dihedral effect is the steady,
straight sideslip flight. Directional stability depends on pnor knowledge of the control
denvatives like c,, and c, . When applying this method (steady. straight sideslip
flight) to find the denvative coefficients, it is essential to set up the maneuver as straight
Ievel flight in order to trirn the data. This means the control ofthe aircraft (speed and
altitude) should not change for at least a short time . This is an effective method as beta
( p ) is not changing rapidly  at least no more than one degree per second and nidder
and aileron have constant track. The rudder and aileron are applied simultaneousfyto
stabilize a sideslip. The rigidbody state is descnbed by:
x = (v. a. q, 0 , h. B, r, p. cp, O)= (7)
13
The first five components of x are called longitud'mal variables and the second five
represent lateraldirectional variables. The rate of change of longitudinallateral with
respect to time ( u, u, IV ) represent the linear velocity dong the axes ( x , y, z) , while
( p, q, r ) are the angular velocity.
The ( 4 , cp. 8 ) are the parameters which determine the aircraft orientation as:
0  illustrates the rotation on yaxis and called the pitch angle
cp  indicates the rotation on xaxis and called the roi1 angle
The values of control effectiveness derivatives c,, , c,, , c,, and c, are
determined fiom the control effectiveness tests. With these values in hand cLp.c , and
c,, are determined fiom the plots of control surface deflections and forces measured at
each of the sideslip angle. The yaw rate derivatives cLrand Cs, are also calculated ffom
data recorded in steady tums initiated with aileron and rudder alone. In this fashion, al1 of
the most important lateraldirectional parameters can be estimated with reasonable
accuracy.
2.2 Longitudinal Stability Derivatives
Al1 stability derivatives are essential though some are more important than others,
particularly those for defining and demonstrating the aerodynamic derivatives of the
aircraft for both longitudinal and lateral motion. The aerodynamic derivatives for aircraft
configuration and crosscoupliig derivatives were achieved on data available in references
[2, 19, 471. The nondimensional stability derivative c, is often zero in trimmed motion,
14
except in cases of asymmetry which is typical in roll, sideslip and tum flight situations. Its
magnitude can dEer considerably and its sign changes in Mach nurnber and in dynamic
pressure. The change of lie coefficient with the change of angle of attack, cL , is always
positive for values of angle of attack below the stall value. The wing contributes up to 85 
90 % of its value.
The nondimensional stability denvative, ch,=, is related to the pitching moment
coefficient with the angle of attack and referred to as the longitudinal static stability
derivative. The vdue of this dimensional denvative is mostly affected by the position of
the aircrafi center of gravity. The nondimensional parameter, C , , ~does not have much
influence on the aircraff motion [ I Il. The dimension less elevator derivatives provide
information about the stability and denvative effectiveness of the control surface and
rnaneuverability of the aircraft. They correlate with geometncal features of the air vehicle
and thereby Facilitate the preliminary design process.
parameter. The elevator effectiveness, which is very important in aircratt design, c,, , is
the most control related parameter in the longitudinal motion of the aircrafi. It can be
determined by maximiting the lift value of the wing and also the range of center of gravity
changes during the flight. Various elevator deflection angles are illustrated in Figure 1.
The pitching moment variation is reasonably controlled up to alpha equal20 deg.
However, as the angle of attack increases, there is a clear loss of elevator effectiveness up
to the complete loss of elevator control above a = 45' [12].
w here:
dotbears the meaning of rate of change with respect to time
F,  force on xaxes
rn mass
u  linear veiocity
r  yaw anylar velocity
q  pirch angular velocity
The equation of rate of change of linear velocity is [ 5 11
w here:
F,  force on zaxes
dot  bears the meaning of rate of change with respect to tirne

w linear velocity
p  roll angular velocity
The acceleration of pitch rate is
where:

1, = I,, 1 , = 1, J, = 1, moment of inenia about y and z axes [22]
The above dimensioniess longitudinal derivatives are the most influentid derivatives for
aircraft motion as shown in Figure 2. The derivative c,= is the L
ei curve dope which is
typically in the range 1 to 8 rad' the linear region of the Liff curve below stall [22 1. The
c dope of the pitching moment curve which detemines the pitch stitfness for stability.
The negative value of c,= required for stability that the pilot can manage and control.
The derivatives c,, and cXhare pitch damping derivatives The cXk is usually of greater
u
magnitude than c, . It has a value in the range from O to 40 rad? The effect of velocity
a
positive, the aircrafi will tend to pitch up with an increase in speed. This results in
increased drag and the greater component of gravity dong the xaxis will then tend to
slow the aircraft dom. In practice, the effective center of pressure of an aircrafi may
move afi with increasing subsonic Mach number, thus creating a tendency to dive
(negative ch,, ), with associated unstable speed behavior [22].The basic aircrafi states of
a
interest are the angle of attack a,and the angular velocity, q. An additionai state is
injected by the actuator of the elevator. The general states and inputs of the plant are as
foIlows 1121:
18
where:
x  state of the plant. q  pitch rate. 6, elevator deflection
y=cx
where:
A, B and C are matrices.
1 9
Chapter 3
Stability Derivative of Aircraft
the c is un desirable as the reversed side force inhibits the pilot's ability to detect
'P
sideslip motion early and makes turns difficult to achieve [13]. This parameter contributes
significantly to the damping ratio of the change in value of the rolling moment coefficient
with sideslip angle, cLB, the socalled dihedral effect. This is very important for lateral
stability and control of the aircrafi. This parameter is very effective in maneuvering
capability specifically when lateral control motion is close to the stall position by the
rudder. The negative value of this parameter is required as it improves the damping ratio
of aircraft motion [L5]. The variation in the yaw moment coefficient with changes in
sideslip angle is considered a static directional or weather stability coefficient. It depends
upon the area of the fin and lever a m . The aerodynarnic value of this parameter is
positive, though the aircraft configuration contribution is negative. For better quality of
the parameter and more stability motion, cSp, should be large, though this might affect
the disturbance fkom side gusts. Change in rolling moment coefficient with change in
rolling angular velocity is referred to as the roll damping denvative, c L pwhich
, can be
determined by the geometry of the wing. The value of this parameter is negative, though
it may become positive when the wing is stalled. The variation in the yawing moment
coefficient with a change in rolling velocity, cNP,is usually negative, although a positive
value is desirable. The negative value of this parameter reflects the smaller damping ratio
of the aircraft disturbance motion. Change in the rolling moment coefficient with a change
20
in yawing velocity, c 4 , has a great effect on the spiral mode, but little effect on dutch roll
mode. The factor contnbuting mon to this parameter is the lifl force fiom the wing. Being
negative or positive depends upon the fin's geometiy. The other denvative parameter for
lateral motion is related to the yawing moment coefficient, cSr. It is cdled the yaw
damping derivative; it has a negative value and is the main contnbutor to damping of the
Dutch roll mode and to the stability of the spiral mode [16].
The directional and lateral control States of the aircraft motion are ;, b, and 4 .
.
Sirnilarly, for lateral motiol the control variables are aileron deflection, 6, and rudder
deflection, g ,as indicated in Figures 3. Change in the side force coefficient with nidder
deflection, c , , ~,is very important when considering an automatic flight control system
whjch uses lateral acceleration as feedback otherwise it is negligible [17].One of the most
.
, is the
important factors for control of the lateral dynamics derivative coefficient cLoa
aileron effectiveness as in Figure 4. This denvative is essential for a low speed flight to
counter asymmetric gusts which tend to roll the aircrafi. The yawing moment coefficient,
c , , ~, which results fkom a rudder deflection, has an equal importance to that of aileron
where:
c  is the coefficient of rudder deflection with the change of side force
a,  vertical tail lie curve dope
r  mdder effectiveness
S,  fin area

S gross wing area
The yawing coefficient with respect to the rudder deflection is
where:
c ,,r  the yawing coefficient which results from the rudder defiection
L , distance between center fin aerodynamic center and the center of gravity of
aircraft
b  wing span
where:
c
=sr
 laterai dynamic derivative coefficient which results fiom aileron
effectiveness
2 ,  distance between fin aerodynamic center and body xaxis
where:
dot  refers to rate of change with respect to time
(x, y, z) refers to x, y, z body axes
L  aerodynamic of rollhg moment
4  moments of inertia on x, y, z axes (i = x, y and z)
p, q, r  body angular velocities
w

p acceleration of roll rate parameter
where:

N aerodynarnic yawing moment
T = (c,tana ,c,)q,S
where:
T t h s t

q,  initial value of dynamic pressure
where:
0 ,cp ,  Euler angular rotation in pitch roll and yaw

q ,r body anguiar velocities
where:
v  Linear acceleration
F,  force on yaxe
P  roll rate parameter
u  linear velocity
The derivatives cLp, cnP , c N r ,and ch are al1 damping moment denvatives.
The roll damping derivative cLpusually lies in the range 0.1 to 0.8 rad? It increases in
magnitude in the transonic region and then decreases again at higher Mach numbers. The
importance of this derivative is that it determine the rollrate performance of a highly
maneuverable aircraft. The yaw damping derivative is csr ; it is always negative and is in
t h e near zero. This variable may increase in magnitude in the transonic region.
The denvatives cXp and ch are crossderivatives that exist because of the
coupling of the rolling and yawing motion of an aircraft; they tend to be difficult to
measure and change significantly with mach number. cspusually lies in the range from 
0.5 to 0.1 rab' and can change from negative to positive values in the transonic region.
.
The derivative of yawing ch , lies between zero and 0.6 rad'. It may Vary considerably
Chapter 4
Mathematical Model of Aircraft
components are longinidinal motion variables, the second five are lateral directional
variables. The longitudinal and lateral motion of an aircrafl about its straight and level
flight path with perturbation can be expressed by two independent equations as follows
119, 203.
where:
F, force on xcoordinate
Tt h s t
c,  coefficient of force with respect to x body axes
IV here:
x  represent subscripts of denvative coefficient on the xmes
y  represent subscripts of derivative coefficient on the yaxes
z  represent subscripts of derivative coefficient on the zaxes
F,  force on zcoordinate
cZr dynamic derivative with respect to p w rate
where:
L  surn of moments on xaxes
c ,, dynamic derivative coefficient of moment with respect to thnist
c  dynamic derivative coefficient of moment resulted fiom alpha
27
c,,  dynamic derivative coefficient of moment resulted from beta
LP
 dynamic derivative coefficient of moment resulted from angular roll rate (p)
cLq  dynamic derivative coefficient of moment resulted from anplar pitch rate (q)
ange
~ 6 a  dynamic derivative coefficient resulted rom aileron deflection

br dynamic derivative coefficient resulted fiom nidder deflection
where:
c
M;  dynamic derivative coefficient resulted from beta rate of change
38
where:

N sum of moment on zaxis
c,
, derivative coefficient resulted from thmst
Csa  derivative coefficient resulted fiom alpha
c,,  derivative coefficient resulted from sideslip
cSi  derivative coefficient resulted from roll. pitch and yaw rate
(where i = p, q and r)
c,.  dynamk coefficient resulted from the change of alpha rate
a
CS.
P
 dynamic coefficient resulted from the change of sideslip rate
c,,  dynarnic coefficient resulted fiom elevator deflection
where:
Y  the rate of change of aircrafi position with respect to yaxis
The acceleration of aircraft position with respect to zaxis can defined
O
where:

0
where:
U, V, w  are the linear velocities
3 0
The angle of attack may be calculated fiom
where:
a  angle of attack
The sideslip angle can be calculated fiom
p = a r c t a n p u)
where:
p  sideslip angle
The initial value of Euler anguiar rotation in pitch can be defined
0 , = arcsin(sin@,sin p , cos@, sin 6 cosp ,)
where:
8 ,,,O,, .p ,,  initial values of angular rotation in pitch. bank and sideslip angles
respectively.
The initial value of EuIer rotation in yaw is
v o = arcsin[(sinO ,sincosp,  cos@ .sin4 ,) i cos8 ,]
where:
v, initial linear velocity
~ ~150)
' 10
P O = 1 . 2 2 ~ 3 e !
where:
hg is the initial altitude
The rate of change of Euler anplar rotations in pitch and roll can be defined
where:
dot  refers to rate of change nith respect to tirne
0 pitch agie
4  roll ansle
The initial v e l o c i ~1s
nhere:
y ,O  "[id values of \;iw and pitch respectively
where:
LT
 dynamic denvative resulted fiom thrust
C ~ P
 dynamic derivative resulted fiom fiom the roll rate
c 4  dpamic derivative moment resulted from the yaw rate
C ~ g
 d@c derivative resulted fiom pitch rate
3 2
The d?namic drivative resulted fiom the control of thnist is
,c, = c,  c,, hoc /2v, )
where:

c, force coefficient on zaus
c
zq
 d>.namicderivative resulted fiom pitch rate
ivhert.:
c ,  damc derivative with respect to thmst
rate
\vhere:
cSp d>namicderivative with respect to roll rate
where:
VI 0 . 0 0 3 4 0
 initial values of Euler angular rotation in yaw*pitch and roll
n i e initial value of pitch rate can be defined
where:

w is the Linear velocity
The horizonta1 flight path angle with respect to vertical plane is
Y H= arctan(+/ k)
The vertical flight path angle with respect to horizontal plane is
When al1 coefficient matrices of the aircraft are obtained ,it shows that longitudinal and the lateral
equations are decoupled and can be solved separately, although the lateral equations depend on the
steady state longitudinal parameters MT. ye and le. In Figure Sb and Figure 5c shows the
horizontal and vertical path angles and y ,.
The Iateraldirectional matrices are shown as:
3 5
Chapter 5
LQR Control of Aircraft
to establish and assume Q and R to be diagonal matrices. It is knom that the LQR has the
folloning p roperties :
sirnplici~in the computation of the gain matrices
Ion order of the compensator .
The problem of the aireraft stabilih may be solved bu ushg fedback application for a11 control
gains and the performance criterion. M e r selecting the above panmeters, the gains can be
computed by matris equations. As indicated above, the plant is described by the statevariable
mode1 [22. 231,
where:
x  is the system state vector
u  is the input vector
A, B, C  are matrices of the state variable mode1
With x(t) E Rn the state, u(t) E Rm the control input, and y(t) tz RP the measured output.
The controls feedback is :
u=Ky
where:
K is an m x p matrix of a constant feedback coefficients.
3 6
Sincc the reguiator problem on& involves stabiliung thc i r c r a f t and inducing good closedloop
timc rtsponscs, u(t) \\il1 be taken as a pure feedback with no au'rili input. In order to obtain
an adcquate response in the control stability of the closedloop qstcm. we wiil sclect a
p r r t o r m c e critenon in the t h e Somain. The goal to drive any initial condition error to zero can
bc achieved by selecting the control input u(t) to miaunke the quadraiic perfomiancc
ivhere Q and R are qmmetric positive weighting matrices. Positive sernidefhiteness of a square
niatris M ( denoted hl p a t e r or equd to zero ) is equivalent to d l its cigenvalues bing
nonnegtive. and also to the requirernent that the q u J d n t ~ cform x 'Mx bc nomegtive for al1
vectors. Theretire, the definite assumption on Q and R guarantee that J is nonnegative ruid Iead
a scnsible muiimizrrtion problem.
By substituting the conirol (64) into (62) the closedlwp synem equations are detennined to be
ivhere:
A, B , C  mauices
The problem is now to select the gain so that 1 is minimired subject to the d~mamicalconstraint
(66)
where:
Q and R are synrneuic positive semidefinite weighuig matrices
1  performance index
xT transpose of the xmatrk
.. *
Selecting the gain value K so J is minunized subject to the dynamical constraints o f equation (73).
37
To simpli the dyimical opthkation problern we convert it h o an equivalent static. that is
a i e r to sohe the problern It is assumeci that we can achieve a positive semidefinitesynmetric
rnatns P and J which may be expressecl as:
.
Assuming that the ciosedloop sysrem is aqmptoticaIly stable so that x(t) vanishes with tirne
Supposing that the x(t) disappears with tirne as it closes to this leads to
If P satisfis the equation (65) then we may apply (66) and obtain the follo~~ing:
Since this rnust hold for al1 initial conditions, and hence for al1 state trajectories z(t) may be
g 
espressed as
A~PP
+ A , + C ~ K ~ R K CQ+ = O
in short, for any fixed value of the feedback, there is a matrix P which m e t s the requirernent for
the equation (73) with a stable ctosed hop system: the performance indes. J . is provided in
terms of P . Thjs indicates that P is independent of the state svhich is the n x n matris.
3 8
Therefore only the initial condition x(0) is applicable to compute the closed loop under the
~nfiuenceof the fedback convol (64) u = Ky . This provides us with an opportunity to
compute the gain values. Now, appiying the trace identie
tr(AI3) = tr(BA) (74)
and their dimensional matrices A and B. we may \rite (70) as the following :
It is concluded that the problem of selecting gain vdue to rninimize (67)subject to the d>mamcal
restrictions (66) on the States, is equivalent to the algebraic solution of selecting the K parameters
to minimite (75) subject to the constraints (73) on the auuiliary matris P .
This problem is solved by using the Lagrange multiplier approach for the constraint to LQR b>
determining the Hamiltonian
H = tr(PX) + tr(gS)
where:

tr is a trace identic
wth S 3 s>mmetric n x n matris of Lagrange multipliers which still needs to be detennined.
Then our consuained optimization problem is equivalent to the simpler problem o f minimizino (77)
without constraints. To accomplish this we need ody set the partial denvatives of H \rith respect to
ail the independent variables P. S and K equal to zero. in sumrnary, for any fked f d b a c k
matris K if there esists a constant, mmetric, positivesemidefinite ma ut^ P that satisfies (73)
and if the closed loop ?stem is stable. the cost J is given in ternis of by (70). This is an
unportant result in that the n x n ausilia matnY P is independent of the state. Given a feedback
matris K. P may be computed fiom the Lyapunov equation (73). Then only the initial condition
x(0) is required to compute the closedloop cost under the infiuence of the feedback control(64).
That is. it may compute the cost of applyng the feedback control u = Ky before w e appiy it. It is
necessary to use this result to compute the gain K tha minimites the LQR by using the trace
identic (74).
CHAPTER 6
Fuzzy Logic Control of Aircraft
6.1 Introduction
One of the latest methodolo@es. FLC is based on the socalled intelligent control
which is defined as a combination of Control Theory and Artificial Intelligence (AI)[28,
[29]. Among many new possible methodologies based on Ai, fzzy logic has been siven
much attention because of its relative simplicity. h e basic idea of a hzzy logic controller.
which is to imirate the control actions of a human operator cm generally be represented
as a collection of "ifthen" rules [;O, 3 11. Unlike binary (yes or no) information f u ~ y
logic is a multivalued logic that permits certain degrees for normal versus slow, high,
and medium set membenhip, which can provide an effective way of overcorning the
problem of inaccurate data in a binary system.
Fu logic's benefits are fault tolerance and the ability to provide accurate
response to ambiguous data. A fLzzy logic controller provides an aigonthm which
convens the linguistic control strategies, based on expert information. into an automatic
control strategy. The methodology of the FLC appears to be very useful when the flight
processes are too complex for systems analysis; LC interprets data qualitatively and
accurately. Generally. fzq logic control provides a clear application of the maximum
input and maximum output (MIMO) systems. In the FLC approach, the membership
functions are defined in order to obtain MiMo structures of longitudinal and lateral
motion adjust.
The application of fuuy logic to engineering problems represents only a fraction of
its real potential. As a method of encoding and using human knowledge, t i i q logic
systems provide the facilities necessary to break through the computationd difficulties
associated with traditional decision support and expert systems. F u z y logic systems have
not yet found their way into aircraft control due to a number of factors that range from
user unfamiliariry with this method and suspicion that there is something fiindamental
wrong with a reasonins system that announces its own imprecision.
Fu logic controllers are basically nonlinear, and general enough to provide the desired
nonlinear control actions throu& carefl adjustment [32]. Fuzzy logic seems to be a
promising method for nonlinear control of the aircrafi. In the context of the general
description of knowledge based systems (KBS) FLC can be introduced to perform a
specific task of closed loop control. The applications of a knowkdge based controller can
be identified as:
a direct expen control system (DECS)and
a supervisory expert controi system (SECS) [33].
With the application of fuzry sets and fzzy operations, it is possible to implement a
design for a fzzy logic reasoning system as a controller. The structure of the fuzzy
controller is illustrated in Fig.6. The control strategy is kept in the form of ifthen rules
within the rule base. These d e s actually represent the mapping of the input variables to
the output actions . The dynamic filters are utilized so as to avoid the introduction of error
and derivatives of error into the output integration. Scaling is performed to retain the
signals benveen the input parameter and the output lirnits for the fuuy process ( rules).
The membership functions produce a clear interface between the linguistic knowledge and
the numerical process variables. The fuuification defines the membership deoree of the
input. The reasoning mechanism combines this information with the rule base and defines
the fzzy output of the rulebase systern. The application of fzzy logic serves to mimic
the control behavior of human action in a direct configuration or otherwise in a
s u p e ~ s o r ycontrol environment. As the mie base indicates a mapping between the
antecedent and the consequent variable. extemal dynamic filters are utilized to introduce
the desired dynamic behavior of the controller ( Fig. 6). The supervisoiy controller is a
secondary controller which augments the existing controller for various conditions. The
advantage of this system is the duplication of the aiready existing system. The fuuy Iogic
controller may be designed according to two different methods:
1. Design the controller From the knowledge available from the domain experts
7 Develop a funy mode1 of the plan? fiorn the measurements, first pnnciples and
expert knowledge, and use this model to design a controller or incorporate this
model in a modelbased control scheme 148,491
Definition 1.
Union: The membenhip ttnction p .x_, of the union A, B is defined for al1 x E W by
Definition 2.
Intersection: The membership function p ,,
of the intersection An B is defined for x
Definition 3.
Complement: The membership fnction y A of the complement of a f k q set is defined
for al1 x EU by
= IPA(X) (82)
which convens the linguistic control strategies based on expert information into an
automatic control strategy. The methodologies of FLC appears very usefiil when
processes are too cornplex for analysis by conventional quantitative techniques. and
attractive sources of information are interpreted qualitatively, inexactly, or uncenainiy.
A funy variable is used to represent an unknown value. A variable symbol can be a single
capital Ietter, a capital letter followed by one or more digits. two capital letters followed
by zero or more characters, or an underscore followed by any character.
6.2.3 Crisp Set
The cnsp set can be descnbed according to the traditional definition of a set in
classical logic as indicated in Figure 6 . Crisp sets have strict membership criteria according
to which an object is either completely included or excluded fiom the set. For example, an
error lies between  1 and +l. In cnsp logic, two logic values are used to define the error
as if the error where above or below zero; thus, the error is positive or negative
respectively. The mapping of the error and the cnsp logic value are illustrated in table 6.1
and Figure 7.
 1 .O 1 NB (negative big)
6.2.6 Implication
The degree of filfilment is used to modifL the consequent of the corresponding
rule. This operation represents the ifthen implication defined as a tnorm, ie . a
conjunction operator (eg. product). Hence the fiizq outputs of the rules become:
6.27 Aggregation
The consequents of al1 d e s are combined into a single fzzy set. The aggregation
operator depends on the implication function used; for conjunctions, it is a disjunction
operator (eg mm):
FLC ouput(u) = ;
max()i (u), p (u))
6.2.8 Defuzzification
Defunification is a process in which fuuy output is converted into cnsp,
numencal results as indicated in Figure 6. Defunification can be considered as an
operator that replaces a fuay set y a representative value.
6.2. IO Membership
Fuuy sets have values between O and 1 that indicate the degree to which an
element has membership in the set. At 0, the elernent has no membership; while at 1, it bas
full rnembership. The function which a s s i g s this value is c d e d the membership fiinction
associated with the f i r q set.
For anaiyzing the parameters of aircraft motion, the linguistic tenns are presented
by membership functions [36]. The position and shape of a fuuy set membership function
depends on the particular application. The variable 'error' in figure 9 ,a senes of
triangularshaped membership fnctions is show. Notice that in this case, the membership
fnctions are overlapping and that the sum is always one. The possible mapping of error
and fzq logic value is s h o w in Table 6.2.
Tablc6.2 Error versus f u q logic values
The fuuy membership functions defined for the consequents rules can also be illustrated
in Table 6.3 . In Figure 8. a senes of membership functions is shown for variable conrrol
action. The large set of consequent membership functions is required to initialize the
controiler by a linear rnapping.
1
0.4 1 0.2 1 PS (positive small)
0.6 0.3 PM (positive medium)
0.8 1 0.4 1 PB (positive big)
1.0 11 1 PVB (positive very big)
In order to apply the rules. one needs to combine the mernbenhip hnctions using logical
connectives such as "and" (conjunction). or "or" (disjunction). The most comrnonly used
conjunction operators are the minimum and the product operators. Usually, the maximum
or sum operator is used for disjunction. The rnapping of error and fiiay Iogic control
action is illustrated in Table 6.3.
The following rules are designed to pave the way to introducing the f k q controller
concept to the control system of the aircrafi. Figure 9 represents the fbzq membership
fiinctions as defned for the consequents of the rules. A large set of consequent
membership functions is required in order to initialire the controlkr by a Iinear mapping.
For the longitudinal motion the mapping includes a pitch angle controller with pitch rate
damping; for the lateral motion, a roll angle holds system and yaw damper. Through the
use of 7 membership functions on the input, we derive 49 mies for each control variable
are derived usine the energy concept (considering the kinetic related to velocity, and
potential related to altitude energy of the aircraf?) [SOI. The knowledge is captured
through the use of If Then rules and linguistic terms like small, big etc. Suppose that the
longitudinal task of the pilot is to control the altitude and velocity of the aircraft. For
example, if the pilot realizes that the aircrafi flies too low and too slow. he will increase
t h s t in order to increase the energy of the aircraft On the other hand, if the aircrafi flies
too low but the velocity is too hioh, then the pilot will increase the pitch ansle. Thrust in
fact increases the total energy, while an exchange between kinetic and potential energy can
be achieved by pitch angle changes (via the tailplane).
The increase of thrust reflects the increase of the total energy, and pitch angle
divides the energy into kinetic and potential. Based on the above definition of the energy
principle, two mle bases for pitch angle and throttle setting can be developed. With the
rule base for throttle setting changes, the rnembership fiinctions indicated in Table 6.4 and
Table 6.5 can be obtained.
52
Table 6.4. Rule base for throttlt membership function setting changes.
Table 6.5. The logic control values of throttle setting membership functions
53
The d e base for pitch angle is illustrated in Tables 6.6 and 6.7
Table 6.6 The mle base for pitch angle changes
PVS 1 ZE 1 N V S
PS 1 PSV 1 ZE
PM 1 PS 1 PVS NVS NS/
PVS
PVB ( P B ( P M PVS ( ZE
PE 1 PVB 1 PB
Table 6.7 The logic coatrol values for pitch angle variations of membenhip
functions
In Table 6.8, a rule base is illustrated with heading error yr = yr ,  yr and the extra
desired heading change AV, as antecedents. and roll angle cornmand as consequent.
Table 68 Rule bases for roll angles and rudder commands.
C hapter 7
Simulation and Analysis
However since alpha measurement are quite noisy. a lowpass filter is used to provide
filtered measurements of a ,and produces the same results as alpha is shown in Figure
17. An integrator (r) was applied to ensure a zero steady state; the output graph shows a
better performance of the fbzzy controller compared with that of the LQR controller as
elaborated in Figure 18. The integrator output is E .
In the fiizzy controller, the angle of attack, the angular velocity q(t), and the
.
integrator are used as feedback gains Ka Kq and KI respectively. The longitudinal task
of the crew is to regulate and control the total power of the air vehicle. If the velocity of
the aircraft is too low, the pilot has to take action to increase the engine thmst in order to
maintain control. On the other hand, if the aircraft nies too Iow but the velocity is high.
then the pilot should increase the pitch angle. The engine thmst increases the total power
of the aircraft, while the exchange (or the difference between potential and kinetic energy)
can be achieved by the pitch angle variations through the aircraA elevator ( tailplane).
Based on the above assessment, the rule base cornrnands for throttle and pitch angle
settings were defined. The labels of the settings are: negative big (NB), negative medium
(NM), negative small WS),zero (ZE), positive srnail (PS), positive medium (PM), and
positive big (PB). These terms are directly related to a senes of membership fnctions
between 1 and 1. The scaling factor with relative terms is transiated into the domain
between 1 and 1. The control actions are the throttle settng and the pitch angle
comrnands. Refemng to Table 6.4. if the velocity error is PM ,the aircrafl motion is too
slow and the altitude error is zero; then the pilot must take action to increase the throttle
change to apply more power as PS to the situation of the flight. The logic control values
of Table 6.4 are indicated in Table 6.5. As clearly illustrated in the simulation results of the
hzzy and LQR controllers for longitudinal motion of the aircraft, the f q control
application gives better results than those of the LQR, as the damping fiequency is much
faster and the situation is stabilized to zero more quickly. Looking at the control of
angular velocity and angle of attack, the results show that the amplitude of the f u q
controller is greater and stabilizes the situation more quickly than the LQR controller
which has a time delay interval. These encouraging results demonstrate the better
effectivenessof the fny controller over the LQR in the application of knowledgebased
control techniques for flight controllers. It is important to point out that for longitudinal
motion, the discrepancy between the linear quadratic (LQR)and tzzy logic controller
(FLC) are very smail as indicated in Figure 16 and Figure 17.
7.2 Lateral Motion
The results of dynamic simulation of lateral motion show far better performance of
the f u u y controller than the LQR.The difference in performance here was greater than
that in the longitudinal simulation. The simulation results clearly indicate the stability and
good performance of multivariable frequencydomain techniques.
The hzqlogic controller provides a transparent interface between the
altitude control of the aircraft and a high level of human action. The EXC requires more
of human intuition than the solving of equations of multivariables for specific formulae.
The performance of the longitudinal and lateral are evaluated by of a time responses under
the conditions of Table 6.4 ,Table 6.6 and Table 6.8..The test of the sideslip ande under
the requirement of the time response shows that the f u q logic controller is more
effective and stabilizes aster than the LQR controller (Figure 18). The results of the roll
angular velocity in dynamic motion without controller, fuuy logic and LQR with
controller are illustrated in Figure 20. In this case, it is very clear that the f k z y controller
has better performance as it has less amplitude and stabilizes faster in less than 3 seconds,
while the amplitude of LQR has a Iittle geater which results in a delay and effects its
performance. Refemng to the simulation results of the yaw rate in Figure 2 1. mdder
deflection in Figure 22 and aileron displacement in Figure 23. They indicate the nature of
the fuzzy controller and its performance provides a better results than the Iinear quadratic
regulator (LQR). The simulation results, show that the application of fuzzy logic
controller gives better performance than LQR controller in the lateral motion of the
aircraft.
Chapter 8
Conclusions
For LQR, and fuuy logic controllen. most of the performance is predetemined
when the strategy and the design is defined. The FLC is a systematic procedure, but fine
optimiPng the t h e responses of the FLC requires a trial and error process. However.
using very restricted, clear guidelines and with a basic understanding of controller
parameter changes, satisfactory results c m be achieved. Aerodynamic derivatives for
configuration at the required flight conditions mut be considered, and great attention
should b e given to these derivatives as they have a tremendous influence on aircraft
motion.
The fuzry logic controller is considered a realtime expert system implementing
the actions of a human operator. The furzy logic controller is more effective than the LQR
as it doesn't require any explicit mathematical mode1 description; instead the actions of a
human operator are modeled. This results in a more efficient controller design, but only if
explicit operator knowledge is available in a suitable form. Finetuning the performance of
the controller is a trial and error process, but by using the provided guidelines and
understanding the influence of the controller parameters, a satisfactory controller can be
developed.
The controt strategy is kept in the form of Ifthen rules within the rule base. They
actually represent mapping of the input variables to the output actions. The dynamic filters
are utilized to input the error and denvatives of error and to introduce the output
integration. Scaling is perfonned to retain the signals between the input output parameter
and the limits for the f t q process ( rules) as defined. The membership fnctions produce
a clear interface from the linguistic knowledge to the numencal process variables. The
fzzification defines the membership degrees of the input to the antecedent fay sets. The
reasoning mechanism combines this information with the rule base and defines the firzq
output of the rulebased system. The application offuay is to mimic the control behavior
of the human action in a direct configuration or otherwise in a supeMsory control
environment. As the nile base indicates a mapping between the antecedent and the
consequent variable. external dynamic filters are utilied to introduce the desired dynamic
behavior of the controller. The supe~sorycontroller is a secondary controller which
augmented the existing controller for various conditions. The advantage of this system is a
duplication of the already existing system.
The performed simulation demonstrated that the hzzy logic controller (FLC) is an
efficient tool in control of the aircraft. More generaily, the FLC is especially usehl and
applicable to al1 nonlinear dynamic systems where the mathematical mode1 is unknown or
partially known.
In this case of the aircraft, the application of the FLC improved the dynarnics of
transient responses and therefore made the control efforts more efficient.
Fig. 1 Farnily of C,a Cuwes for Various 5, [IO]
M ~ c hN u n b e r
Fig.2 variation of Longitudinal Stability Coefficients with Mach Number 1141
ANGLE OF A T TACK  OEG
Fig. 3 Loss rudder effectiveness due to combhed variations of angle of
artack and sideslip [15)
60
ANGLE O F ATTACK  OEG
I_
' O , 50"
U
 0.4
1
Fig. 5.2 Subsonic cross derivatives for a swept wing fighter aircraft a M=0.7 [12]
Fig. 5.3 Variation or crosscoupling derivatives with the angle of attack 1331
Fig. Sa Body system ofaxes and related angles, positive directions indicated by arrows
Fig. Sb Vertical fight path angle,. vertical plane
5
c
Fig.8 Fu controller
Fig. 9 Output membership functions
Fig. 10 Input membership functions with input values
Aed Product lJ

*
0.8 ' I t
10
I
15
O 5
Tirne (sec)
Fig. 16 Longitudinal motion (6.  Elevator deflection versus Time)
3r
1 
1 ",
 Futzy conuoller
 LQR conuoller
I
O
O 5 IO 15
Time (sec)
Fig. 17 Longitudinal motion (aF Filtered measurement of angle of anack versus T i r
1 
0.8 
\
0.6
1  Fuzzy conuoller
 LQR conuoller
. withou~ccntrollcr
 .

   . 
I
0.2 1
I
I
' /
I
J
I
1
0.4  \
1
I
1
i
0 6  I
1 I 1
O 5 10 1s
Tirne (sec)
O 5 10 1S
Time (sec)
Time (sec)
 +mut controller
1O
Time (sec)
Fig. 22 Lateral motion (Rudder deflection versus Time)
Fig. 23 Lateral motion (Aileron deflection venus Time)
References
Luenberger, D. G., "An Introduction to Observers," IEEE Trans. On Automatic
Control, pp.596603, December 197 1.
Grafion S. B. & Libbey C. E., "Dynamic Stability Derivatives of Twin jet Fighter
Model for Angles of Attack corn  10' to 1 10~,".NASA, TN D609 1, 1971.
Rom,J.. "High Angle of Attack Aerodynamics," SpringerVerlag, 1992.
Visbal, M. R."Dynamic Stall of Constant Rate Pitching Aerofoil. J. Aircrafi,"
27 (may), 1990.
Huang, Chien Y., et al., "Analysis and Simulation of Nonlinear Control Strategies
to Hgh AngleofAttack Maneuvers," Proceeding of AIAA Guidence, Navigation,
and Control Conference, New Orleans, Lo 199 1.
Favilla.. J., ''Fuuy Traffic Contro1:Adaptive Strateies," Second IEEE International
Conference on Fuzzy Systems, 1993.
Anderson, B. D and Moore. J. B., "Linear Optimal Control". Prentice Hall. 1992.
OrlikRuchemanri, K. I., "Aerodynamics Coupling between Lateral and
Longitudinal Degrees of Freedorn," AIAA journal, Vol 15,No. 12, Dec. 1977.
Pontryagin. L .S.." Ordinary Differentiaf Equations," Addison Wesley, 1962
OrlikRuchemann, K. J., Hanf, E S . , Laberge, JG. " Direct and CrossCoupling
Subsonic Moment Derivatives due to Oscillatory Pitching an Yawing of an Aircrall
like Model at Angles of Attack up to 40' in AMES ."NRC NAE LTRUA 38,
Nov. 1976.
Johnson, H. 1.. "Flight Testing Aircraft for Longitudinal Maneuvering
Characteristics," Vol. 2, AGARD Flight Test Manual, Pergamon Press, New
York, 1959.
Nelson, R.C., "Flight Stability and Control Automatic Control," McGrawHill
Book Company, New York, 1989.
Babister ,A.W., "Aircraft Dynamic Stability and Response," Pergamon
Press. 1980
Mclean D.,"Automatic Fiight Control System" PrenticeHall Intemational. 1990
Coe, P.L., Graham, AB..and Chambers,J.R., "Sumrnary of Information on Low
Speed LateralDirectional Derivatives Due to Rate of Change of Sideslip Angle,"
NASA TN D7972, Sep. 1975.
Hancock. G.J. & MAbey, D.."Unsteady Aerodynamics of Controls," AGARD CP
465,1978.
Atzhorn D.,and Stengel, R. ,"Design and Flight Test of a LateralDirectional
Comrnand Augmentation System," JGuidContDyn., VoI.7, No.3. MayJun 1984.
Hancock G .J.. "An Introduction to the Flight Dynamics of Rigid Aerosplanes,"
Eiiis Honvad Limited 1995.
"Unsteady AerodynamicsFundamentals and Applications to Aircrafi
Aerodynamics.," AGARD CP 386,1985.
Etkin. B.. "Dynamics of FlightStability and Control, ( 2nd Edition) ," John Wiley
& Sons. New York 1982.
Selson. R. C.,"Flight Stability and Automatic Control," McGrawHill, 1989.
Bnan, L. Stevens, Frank LLeewis, "Aircrafl Control and Simulation7" 1992 by
John Wiley & Sons, Inc.
Blakelock, J .FI.., "Automatic Control of Aircraft and Missiles (2nd Edition),"
John b'iley & sons, Inc, New York 1991.
Smith, GA.. and G. Meyer, " Aircraft Automatic Flight Control System with
Mode1 Inversion ",Journal of Guidance, Control and Dynamics, Vol 10, No.$
MayJune 1987.
Mclean D., "Automatic Flight Systems," PrenticeHall Intemationai, 1990.
Isermann. R, LactimannJCH., Matko, D..,"Adaptive Control Systems ."Prentice
Hall, 1991.
Huang*Chien Y.,"Multivariable Control Law for FlatTum Strafnng Maneuver by
a Supermaneuverable Aircraft" Proc. of AIAA Guidance, Navigation, and Control
Conference, Aug. 1990.
Friedland Bernard, "Control System Design, " McGRawHill, Inc. 1986.
86
Appendix : A
A. 1 Characteristics o f aircraft
The aerodynamic derivatives of hypothetical aircrafi configuration based on data
available in references [2, 5, 103
Table A.1.1
90 0.0 ( radsec)
r
1t 7.85 cm)
 
C,/C 045
A.2 Aerodynamic and control derivatives
The parameters of the control derivatives are based on data avaiiable in references [2,
471.
Ta bie A.2.1
Parameter
Appendix: B
* *
* THE FUZZY LOGIC PROGRAMMING FOR AIRCRAFT MOTION *
* T
kdefine pi 3.14159
float calbif(float[],float.float,int);
float rnemberfinction(int,int.fioat,float,float);
float rnemberJnctionG(int.int.float.float,fioat);
float fzzycontrol(fioat[]);
float ucrisp(int.float[J,float[],float.float);
void rnain(void)
{
float x[jo].xa[jO],m[jO],k 1[SOI, W[50],W [50].k4[5O].u.t,h.gO.g 1 .H:
int j,i,l,Nooceqs;
FILE *@;
if((@f0pen("gs2~'.~'~vb"))=NULL)
{
printf("cannot open directory fiiei");
exit (1);
1
/*initiaiizations/
X[ 1 ]=. 1; x[2]=0; x[3]=0; g0=2; g 1=. 1 ; H=S;
u=O; FO; h=.OOl;
Noof_eqs=3 ;
for(j=Oa<=3 1 OOj*)
x[l]=gO*x[l]; x[2]=glfx[2];
/*The differential equations of Aircrafl by the RungeKutta methodi/
x[ 1]=x[ l ]/go; x[2]=?c[2]/g 1;
for(l=0;1<=3I Oit+)
{
i@1' 10)
Fprin~Q,t'.4%d=f%f%f%fl;\n".l.t,u.x[l]);
}
for(+ 1:i<=Nooeqs;i*)
xn[i]=s[i];
1
for(+ 1 :i<=Noof_eqs;i*)
{
1
for(+ 1;i<=Nooeqs;i++)
C
~a[i].cn[i]k 1 [il/':
for(+ 1;&=Noofeqs;i*)
{
k2[i]=calbif(xau. hi);
i
for(+ l :i<=Nooqs;i++)
{
xa[iIq[i]+ k2[i]/2;
}
for(i= l ;i<=Noofeqs;i*)
{
k3 [i]=caidif(xa~~,h,i);
1
for(i= 1 ;i<=Noofeqs;i*)
{
xa[i]icn[i]+k3 [il;
1
for(i= 1 ;i<=Nopfeqs;i*)
{
k4[i]=caldif(xa,u, h, i);
iSSTATESPACEEQUATION OF Aircraft*/
float caldiqfloat xa[],float ufloat h,int i)
{
fioat a,b.k[SO];
k[l]='ca[2]*h;
a=9.8 *sin(xa[ 1])+cos(xa[ Il)*(xa[3]0. X*pow(xa[2].?)*sin(xa[1]))/ 1 5:
b=OS *(4/3pow(cos(xa[ 1]),2)/3);
k[2]=h+a/b:
k[3]=h*( 1 OO*xa[3]+100*u);
return k[i];
i
J
fioat rnf122J;
Mi]=O;
if(i<N/2)
{
iRxdC0)
iKxd<=cl)
M O ] = 1;
else
{
if(((iN/2 I )*w<=xd)&&(xd<=(iNl2)*w))
mai]= 1((iN/2)*w+xd)/w;
else iK((iN/2)* w<xd)&&(xd<=(iN/2+ 1 )*w))
mfli]= 1+((iNE) *wxd)/w;
else
mfli]=O;
}
1
else
mqi]=O;
eIse if(i=N/2)
if((wcxd)&&(xd<w))
{
if(xd<=O)
1nfl3/2]= 1+xd/w;
else
1
else
mfli]=O:
1
retum mfli];