You are on page 1of 5

Automatica, Vol. 20, No. 1, pp. 103 107, 1984 0005-1098/84 $3.00 + 0.

00
Printed in Great Britain. Pergamon Press Ltd.
1984 International Federation of Automatic Control

Brief Paper

Application of Nonlinear Transformations to Automatic Flight


Control*

G. MEYER,t R. SUJ; and L. R. HUNT§

Key Words--Aerospace control; air-traffic control; automatic control; control applications; control system
design; Lie algebra; linearization techniques; mathematical system theory; nonlinear control systems;
nonlinear transformations.

Abstract--This paper presents the results of an application of


transformations (from nonlinear to linear systems) to the design of REFERENCE
a helicopter autopilot. The topics covered are (1) a review of the INPUT Yo - + y x

transformation theory and its use in the design approach, (2) a


ey
description of the helicopter mathematical model, (3) the
construction of the transformation from the nonlinear system to m
~=Ay+Bv
the linear system, and (4) a discussion of the system performance. ;¢=f+Zui9
i=1
i

gv
1. Introduction V0 Vo+ Gv
LETTHE state equation of the process to be controlled by given by

~(t) = h(x(t), u) (1)


L I N E A R SYSTEM REGULATOR TRANSFORMATION
I NONLINEAR
where the state x e X c ~", the control U E ~ m. and h is c¢~ on A N D INVERSE PLANT

X x 9t". Under what conditions on h does there exist a


FIG. 1. Design concept.
transformation Tmapping this nonlinear system to a controllable
linear system? Equivalently, are there new coordinates (in state
and control variables) so that system (1) is a controllable linear
system, which we can assume is in canonical form (Brunovsky, transformation involving controls) and applied to the nonlinear
1970)? plant in this way. States x are read from the plant and mapped by T
Necessary and sufficient conditions under which transform- to linear states y. The error ey = y - Yo is passed to the linear
ations from nonlinear to linear system exist are known (Su, 1982; regulator, and a control 6v is found to send er to zero. Disturbances
Hunt, Su and Meyer, 1983a). These are local results (near a point), and variations in plant dynamics are handled in this way. The
and global results are also available (Hunt, Su and Meyer, 1983b). controls v0 and 6v are added and applied to the plant, again
For time varying systems the theory has been developed (Hunt through T - t. In summary, the difficult tasks of finding the 'open-
and Su, 1981). loop' and regulator controls are performed on the linear system,
Transformations for block triangular systems have been making possible an easier design. In general, both the model servo
considered (Meyer and Ciolani, 1975, 1980). Other interesting. and the regulator will have dynamics and nonlinearities such as
results concerningtransformations of nonlinear systems are found limiters. Thus our control actions may not be smooth (c~®).
in the literature (Krener, 1973; Brockett, 1978; Jakubczyk and However, in (1) h is taken to be c ~ but we do not require that u be a
Respondek, 1980). c ~ function of t.
Suppose that our nonlinear system (1) is transformable to the The design approach of this paper and of other references
controllable linear system (called the canonic model) (Meyer and Cicolani, 1975,1980; Meyer, 198 l) has been applied to
several aircraft of increasing complexity. The completely
= Aoy + BoY (2) automatic flight-control system was first tested in flight on a DHC-
6. This test required the aircraft to operate over a large portion of
in Brunovsky form with Kronecker indices xl, K2..... x,,. Then its flight envelope, and the system performed well despite
our design scheme for automatic control is based on Fig. 1, which disturbances and plant modeling errors (Wehrend and Meyer,
emphasizes the model following character of our approach. 1980). The method was also used for the Augmentor Wing Jet
In Fig. 1 Y0 and Vo are states and controls, respectively, STOL Research Aircraft and successfully flight tested (Meyer and
generated by the model servo in response to the reference input. Cicolani, 1980). Incorporation of pilot inputs into the design
The 'open-loop' control Vo is fed through the inverse technique was accomplished (Wehrend, 1979). The approach was
transformation T - t (actually through that part W of the inverse also applied to control an A-7 for carrier landing and testing in
manned simulation (Smith and Meyer, 1979, 1980). In this paper
the design method using transformations is illustrated by an
application to a helicopter.
* Received 23 November 1982; revised 1June 1983. The original Section 2 contains an overview of the theory of transformations
version of this paper was not presented at any IFAC meeting. This from nonlinear to linear systems. A description of the helicopter
paper was recommended for publication in revised form by plant and the construction of the W T map are the subjects in
associate editor V. Utkin under the direction of editor H. Sections 3 and 4, respectively. In Section 5 we describe the system
Kwakernaak. performance of a detailed simulation.
t Mail Stop 210-3, NASA Ames Research Center, Moffett Field,
CA 94035, U.S.A. 2. Transformations
:~Department of Electrical Engineering, Texas Tech University, We examine system (1) and assume that h(0,0)= 0. This
Lubbock, TX 79409, U.S.A. assumption is not essential to our theory (Su, Meyer and Hunt,
§Department of Mathematics, Texas Tech University, 1981), but it does allow us to reduce the notation involved. For
Lubbock, TX 79409, U.S.A. single-input systems like (1) it is proved (Su, 1982) (and this proof

103
104 Brief Paper

can be extended to the multi-input case) that a necessary condition satisfying the hypotheses of this theorem. As we shall see, our
for such a system to be transformable to a controllable linear mathematical model of the helicopter is approximately block
system is that there are new state and control coordinates (also triangular.
called x and u) so that the system becomes In Section 4 of this paper we construct the WT map for the
helicopter model. Introducing a slight change of notation to be
2(t) = [ ( x ( t ) ) + u(t)g(x(t)) (3) used in the remainder of this paper, Twill be (7"1, T2..... T,,),instead
of(T1. T2..... Tn, T.+ 1, T,+2 . . . . . T.+~).and Wwill be theinverse of
7'.+ 1. T,,+2..... T.+ m with respect to u for fixed values of x. This
w i t h / a n d g %'~ vector fields in a neighborhood of the origin in .~" agrees with Fig. 1 where the emphasis in the forward
and ((01 = 0. Hence we can assume that system (1) is transformation is in taking x states to y states and in the inverse
transformation taking v controls to u controls. This is entirely
.¢z(t) = ] ( x ( t ) ) + ~ ui(t)gi(x(t)) (4) compatible with the general transformation theory because
i 1 Tt, T 2 ..... T, are functions of x (and not of ut and only
To+ l, T.+2 . . . . . Tn+ m depend on u. Thus we write 3' = T[-¢) and u
with f(0) = O. = W ( x , v).
The type of transformations of interest to us are described in Since v is the standard notation for velocity, we will introduce a
the following way. We want T = I T 1 , T2 ..... T.T.+I,T.+2 ..... new control variable for the canonic model. We now discuss the
Z,+~) to map from .~¢"+~(xl.x2 . . . . . x,,, ul, u2 . . . . . u,.) space to mathematical model of the helicopter.
~"+~(Yl,Y2 ..... y., cl, c2..... c,.) space with (7"1, T2..... T.)
taking an open neighborhood of the origin in x space to a similar 3, T h e p l a n t - - a helicopter
neighborhood in y space. We want (7"i, T2..... T.) to map the The helicopter will be represented by a rigid body moving in
origin to the origin, Tto be one-to-one, T i to be c g ~ T~, Tz . . . . . 7". three-dimensional space in response to gravity, aerodynamics,
to be functions of x l , x2 . . . . . x . only with nonsingular Jacobian and propulsion. The state
matrix, and T,, + i, T,,+2 . . . . . T.+,. to be functions o f x and u with
nonsingular Jacobian matrix with respect to u = (Ul, u z , . . . , u,,)
x =(r,v,C,o))'E X c,~3 x .~3 x S 0 ( 3 ) × :~¢~ (5)
when x is fixed. Most importantly, we require that T maps the
nonlinear system (4) to the linear system (2). The variables where r and v are the inertial coordinates of body center-of-mass
T1. T2 . . . . . T,, ( = 3q, Y2 . . . . . y., respectively) are the new states and position and velocity, respectively (v should not be confused with
T,, + ~, 7", + z . . . . . 7". + ~ ( = y . + 1. Y. + 2 . . . . . y . +~, respectively) are the the control v used in Sections 1 and 2) and C is the direction cosine
new controls. matrix of the body-fixed axes relative to the runway-fixed axes
Before proceeding we need the following definitions. Given (8~ (taken to be inertial). The attitude, C, moves on the Lie group
vector fields f and g on ..~'" we define the L i e b r a c k e t of f and g SO(3). The body coordinates of angular velocity are represented
by o9. Here the prime denotes transpose.
~g , ~[ The controls

[ j : g ] = ~.x ! - ~ - g (6)

where ?gfi?x and 8 f / S x are Jacobian matrices. Similarly we take where u M is the three-axis moment control; that is, roll cyclic and
pitch cyclic, which tilt the main rotor thrust; and the tail rotor
( a d % g) = g collective, which controls the yaw moment; and u e is the main
(ad[[ig) = If, g] rotor collective, which controls the inain rotor thrust.
The effectively twelve-dimensional state equation consists of the
(ad2j2g) = [ f [./;g]] translational and rotational kinematic and dynamic equations

= jr(x,u)
(adkf g) = [f, (ad k - i f g]. (7)
d = S(og)C
A set oft8 ~ vector fields {fl,.[2 ...... f~} on ..~"is involutive if there d)=fM(x,u)
exist ~ functions 7~jk(x) so that
where j.F and j.M are the total force and moment generation
processes, and (x, u) are defined by (5) and (6).
[[i,[i](x) = ~ 7ijk(x).£(x), 1 <~ i,j <~ r,i ~ j .
k-1
Consider next the transformation of (7) by means of the
mappings u = W ( x . v) and y = T(x).
We define the sets
4. T h e W T - m a p
C = { g l , [ f g l ] ..... ( a d " ' - i f g l ) , g 2 , [ ~ g 2 ] ..... (ad~-l~g2) ..... We want to show that (7) satisfies the conditions for
transformability shown in Theorem 1. First, we put these
equations into a simpler form.
g,,,, [f, g,. ]..... ( a d " - if g,.)}
In general, the moment generation p r o c e s s f M is invertible with
respect to the pair (cb, uU), and for the restricted class of maneuvers
C~j = {gb [./,gt ]..... (ad K~ Z f g l ) , g 2 , If, g2] . . . . . (adK'- 2fgz),
being considered in this experiment (i.e. no 360 ° rolls), f r is
invertible with respect to the pair (b3, uP). Thus, in the present case
.... g,.,[.[~g,.] . . . . . (ad Kj 2fg,.)}, for j = 1,2 ..... m.
for the set of angular and vertical acceleration commands
restricted to the set
The following result (Hunt, Su and Meyer, 1983a) gives
conditions under which a transformation of the type we consider Uo = {(dg,r)3):l~bi[ ~ 0.5 rad/s, i = 1,2,3,1f'31 ~< 0.5gl (8)
exists.
a function h u : X x U60---~ U can be constructed so that if
T h e o r e m 1. The system (4) is transformable to the system (2),
where the state variables xl, x2 .... , x, are contained in some open u u = hU(x, (690, b30)) (9)
neighborhood V of the origin in ~", if and only if on V
1. The set C' spans an n-dimensional space at each point. then
2. The sets (~j are involutive for j = 1, 2..... m.
3. The span of each t~j is equal to the span of Cj c~ C. (b = ( b 0
We remark that block triangular systems (Meyer and Cicolani, (I0)
1975, 1980) are an interesting subset of the class of all systems b3 = v30-
Brief Paper 105

If (tbo, ,33o)are chosen to be the new independent control variables Showing zero entries in block form and specifying the rows
toreplacethenaturalcontrolsu = (uu, ue),thenthestateequation containing possible nonzero entries, computations yield
(7) becomes the following:

t~=v

(f)l)
=f°(r'vt'v2`C)+ef1(r'v'C'0)'(tb°'i)3°))(ll)02

~a = ~33o g, = ~6th,
d = S(~)c

6) = 6)o
I°l
where ~ = 1, and f~, representing parasitic effects, is such that
fl(r,v,C,O,(O,O)) = O.
Ignoring these parasitic effects, which will be handled by the
regulator, we move system (11 ) into the form of (4). We let ~t, ~2,
[f, g l ] = 1117(X ,--7th,
~l,,s(X)[ ~ 8 t h
[° t
1127(X ,-Tth,

and 4'3 be local coordinates for SO(3), with ~1 and ~2 L o/


corresponding to the two-dimensional tilting of the rotor disc, Vo-]
which is in turn controlled by the cyclic pitch. Moreover, ¢~a is
[f, g3 ] = - I l l '--9th,
chosen in such a way that ~3 = 0)a and ~b3 will be called Ea(~b ) in
later discussions. Setting f o = (fo,fo) we find /o/
r I = U1 0

r 2 = V2 121a(x)l ,--4th
r 3 = v3 121s(x)l ~ 5 t h ,
Vl = f°trl, r2, r3, Vl, I)2, (~1, (~2, ¢ 3 ) [f,g,] = - 3rd, (ad2f, g l ) = 0 .
i
V2 = f°(rl,r2, r3, v1,v2,¢l,(02,(~3) 12:(x)[ ~ 7 t h
v3 = b3o (12) 1218(x)1 ,--8th
~b I = hl(q~l,t~2, t ~ 3 , 0 ) l , 0 ) 2 , 0 ) 3 )
0
0
q~3 = tb3 122,(x)1 ~ 4 t h
122s(x)[ ,-- 5th,
(D2 : (D20
(ad2f, g2) = 0
0) = (b3o
1227(x)[ ~ 7 t h
We define xl = r~, x2 = r2, x3 = r3, x,, = vl, x5 = v2, x6 = va, I~atX)[,---8th
X, = ~1, X8 = 1#2, X9 = (#3, X l o = LOI, X l l = 0)2, X12 = 0)3,
ut = O51o, u2 = ¢b2o, u3 = ¢b3o, u, = b3o. Also 0
0 ~--
-X4

X5
(ad2f'g3) I/23~(x)l--4th
I1244(X)],__4th
X6
/iO(x I X2,X3,X4, Xs,X'~,X8,X9
J20(XI, X2, X3, X4, X5, X'r, X8, X9)
-131x(x)l ,-lst -1321(x)-I , - l s t
f= 0
hI(X%X8,X9, Xlo,XI1,XI2)
03) 1312(x) ,-2nd 1322(x) [ ~ 2 n d

h2(xv,Xs,X9, Xlo,Xll,Xl2)
131~(x) ~,--4th /32~(x) ,--4th
0)3
0 (ad3f, ga) = 1315(x) *--5th, (ad3f, gz)= /a250(x) ,--5th.(15)
0
0 13l~(x) ,--7th la2,(x) ,--7th
_131;(X~ ,-8th _132~(X)~ ,--8th
and

Fo- -o-I 0- 0
IO OI
Here the functions lljk(X) are not explicitly displayed in order to
0 0
IO OI
conserve notation, but they can be easily calculated.
0 0
In our model of the U H - I H helicopter, the matrices
'0 OI 0 0
0 OI 0 0
gl = 0 ,g2 = OI, g 3 = 0 , g4 = 1 • (14) [,,.(x) ,,.(x) 1 1 [,.,(xl
0 Ol 0 0 /1Is(x) 112s(X)J'[_1215(x) 122s(x)d'L1312(x) la22(x)J
0 OI 0 0
0 OI 0 0 are assumed to be nonsingular. Thus
1 O[ 0 0
0 1 0 0 ~r = {gl, [f, gl ], (ad2f, gl ), (ad3f, g, ), gz, [f, g2 ],
_o_ _%1 l_ o~ (ad2f, g2), (adSf, g2),g3, [f, g3 ],g4, [f, g4]}
106 Brief Paper

is a linearly independent set of vector fields. Moreover (ad2t~ g3) change the natural representation [(7)] to the canonical
and (ad2tlg,,) [and hence (ad3I~g3) and (ad3fg4)] are linear representation are approximately as shown in the figure. The
combinations of I f g~ ], If, gz ], (adZ.(,,g~ ) and (adZf, gz). Hence the coordinate change, y = T(x), is given by [see (11) and (16)]
Kronecker indices are ~:~ = 4, x2 = 4, K3 = 2, and x,, = 2.
The sets yl = (y31,y~), = (rl,r2)'
y2 = (y~,y~), _ (vl,rz)'
C1 = ( : = [g,, ELg~ ]. (adZ I;g, },g2, I f g2 ], (adZf, g2),
y3 3 3 3 3 (21)
= (YbYz,Y3,)4) = (J11],.I21],E3(~/), r3)'
g3, [.[~g3 ], (adZl; g3 },g4, [,1;g,~], (adS.I; g4)}
_ ~g], g2, g3, g,*~ 24 =
• 4 4 v....
I,Y l , Y 2 , _ 3 , Y 4 )
= [('~J'O
~-~CO, CO3,1)3]I'
are involutive, and ffi - Cj c~ C for j = 1, 2, 3, 4. Therefore the
hypothesis of Theorem 1 are satisfied and our nonlinear system is where c o s ¢ = c11/c, sin~O = c l z/c, c = x/ (c~ l + c22), (cifl - C
transformable to a linear system. and the prime denotes transpose. In the notation of Section 2
The transformation can be computed at this point, but we first
return to the notation of(11 ) (still assuming that e, = 0) and make ( Tb Tz, T3, T4, Ts, T6, T, , Ts, Tg, TI o, Tl l , TI z )
an additional modification. 1 2 3 4 1 2 3 4 3 4 3 ,4
= (Yl,YbYt,YbY2,Y2,Y2,Y2,Y3,Y3,Y,*,Y4)
The function f 0 is invertible with respect to the pair,
[(G,bz,E3(~)),CI, in which E3(t)) is an elementary rotation and y~, y5, y~, y5 are the new-controls.
about the runway z-axis, representing the heading of the The controls variable change, u = W(x, yS), is defined in two

!t
helicopter steps

E3(t))=
(cos s
-sin~O costp . (16)
,~I") v~
tb= 0 0 1 Y~ (22)
0 0
~,5
. 3
If the horizontal acceleration c o m m a n d s are restricted to the set ~3 = Y]
U, = llbl,b2):Jb~l ~ 0.5g, i = 1,21 (17) u = hU(r, v, C, e), (tb, i'3 )1. (23)

then a function hFL:3~3 × ~3 × Ub × S0(2)---, S0(3) can be The effects of the various approximations made in the
constructed so that the helicopter attitude given by construction of these transformations are to be controlled by the
regulator described in the following section.
Co = hF(r, C, (;o, E 3 ( ~ ] / O ) ) (18)
5. The system performance
results in the commanded acceleration The experimental flightpath is defined by a set of way points,
segments of lines and helixes, and a speed profile, as shown in Figs 3
b = Vo. (19) and 4. As can be seen from Fig. 3, the flightpath is a closed curve.
The time dependence is shown in Fig. 4.
Equations (9) and (18) are the trim equations of the helicopter (7) The experiment consists of automatically flying this trajectory.
without the parasitic effects [r = 0 in (11)1. That is, for a given This exercises the system over a wide range of flight conditions.
motion Ir(tj, E3OP(t))], t >1 O, the corresponding trim state and The helicopter is taken from hover (way point 1 in Fig. 3) to high
control may he computed as follows:

r 0 -- rlt)
v0 = ~(t)
Co = ht(ro, Vo, J)0(t), E3(~Oo(t)))
(20)
coo = q(('o(t)C'o)
& o - d~o(t)

Uo = hU(ro, Vo, Co, COo,(6)1],/:30))


.a
u.
E
E / 2s0 m/s
where the function q extracts ~o from S(~o) = CC'. The required
time derivatives in (20) are computable provided that the motion oz
(r, E3) is generated by the strings of integrators shown in Fig. 2, - 2438 m
where the overdot represents a scalar integrator, and y5 E ~,4 is an
independent control variable. The system shown in Fig. 2
(Kronecker Indexes {(4, 4, 2, 2}) will be taken as the canonical
model of the helicopter. The transformation and feedback that
600
3o7\_
y5 y4 y3 y2 yl ~.zoo
E

I=- @ ~. = 8 r1
400
..J
u.
300

200
>
~ =E 3
101]

~.- ~ Cr 3
FIG. 3. Experimental flightpath shown in horizontal and vertical
FIG. 2. Canonic model {4,4,2,2}. planes.
Brief Paper 107

a; ~ a~ ACCELERATION
.2r jf ' ~ I tl I
51
r ~ ANGULAR ACCELERATION

-.2L
3 ~
~ ,,J--
VELOClTY~
,,/ ,", -5
5~_.
L uM MOMENT CONTROLS
ol-------i-----~-.~'~:-----~---
E ~ ~" I
-3 L SIN~k* HEADING " ~ - J POWERCONTROL . .

' r :"" ;Z
0p.-.,.- ./---.-.,....,r.~-.t - - r - . - - - -,,.~.- -~.7,,
'Z 0"r,........
. . . . •.... -+-,--.+ ........
+ " '--+
-- -5 I I I I I I
0 100 200 300 400 500
RELATIVE TIME, see
~° ~
i i i i i FIG. 6. System performance--natural variables.
-5 100 200 300 400 500
RELATIVE TIME, sec

FIG. 4. Time dependence of the coarse command. performance of the automatic control system in a detailed
simulation on the flight computer is summarized. We feel that the
successful application of the design technique using nonlinear to
linear transformations in the UH-1H simulation illustrates the
speed [ 150 ft/s ( ~ 46 m/s) ] accelerating, turning, and ascending strength of the method.
flight. This input to the system is coarse, with a variety of
discontinuities. Acknowledgements--This research was supported by the
The results of a detailed simulation, in which the complete National Research Council at NASA Ames Research Center and
control code was implemented on the flight computer to be used in by the NASA Ames Research Center under the Intergovernmental
the flight test, are summarized in Figs 5 and 6. For the data Personnel Agreement Program and the Joint Services Electronics
presented, the helicopter was flown manually to the point marked Program at Texas Tech University under Office of Naval Research
by a star in Fig. 3. There the automatic system was engaged. It contract N00014-76-C- 1136.
takes about 500s for the helicopter to go once around the
flightpath.
Unlike the coarse accelerations in Fig. 4, the model References
accelerations are smooth, as is the vertical velocity, v30. The Brockett, R. W. (1978). Feedback invariants for nonlinear systems.
second panel in Fig. 5 shows the effects of the neglected parasitic IFAC World Congress, Helsinki.
terms on acceleration. The acceleration errors are quite small; less Brunovsky, P. (1970). A classification of linear controllable
than 0.05g. The regulator controls these effects by means of systems. Kibernetika (Praha), 6, 173.
position errors. The resulting horizontal error is less than 6ft, Hunt, L. R. and R. Su (1981). Control of nonlinear time-varying
while the vertical error is below l ft. The speed error, systems, 20th IEEE Conference on Decision and Control, San
ev = Ilvll- lit011, is below 1 ft/s. Diego, pp. 558-563.
The first panel in Fig. 6 shows the required angular accelerations Hunt, L. R., R. Su and G. Meyer (1983a). Design for multi-input
for exact model following. The correspond.ing moment controls systems. In R. W. Brockett, R. S. Millman and H. J. Sussman,
are um and ue. As may be seen fi:om the figure, the natural controls (Eds), Differential Geometric Control Theory Conference.
are well behaved. Birkhauser, Boston, vol. 27, pp. 268-298.
In summary, the performance of the system is good. Future tests Hunt, L. R., R. Su and G. Meyer (1983b). Global transformations
will exercise the system with more taxing trajectories. of nonlinear systems. IEEE Trans Aut. Control, AC-28, 24.
Jakubczyk, B. and W. Respondek (1980). On linearization of
6. Conclusions control systems. Bull. A cad. P olon. Sci., Ser. Sci. Math A stronom.
The theory of transformations of nonlinear systems to linear Phys., 28, 517.
systems is applied to design an automatic flight controller for the Krener, A. J. (1973). On the equivalence of control systems and the
UH-1H helicopter. We give a description of the helicopter linearization of nonlinear systems. S l A M J. Control, 11,670+
mathematical model and show that it satisfies the necessary and Meyer, G. (l 981 ). The design of exact nonlinear model followers.
sufficient conditions for transformability. The mapping taking the Proceedings of1981 Joint Automatic Control Conference, FA3A.
nonlinear model to canonical form is constructed. The Meyer, G. and L. Cicolani (1975). A formal structure for advanced
automatic flight control systems. NASA TN D-7940.
Meyer, G. and L. Cicolani (1980). Application of nonlinear system
inverses to automatic flight control design--system concepts
and flight evaluations. AGARDograph 251 on Theory and
3 3 Applications of Optimal Control in Aerospace Systems (P. Kent,
YlO Y20
"2gr t "~'~ l" MODEL ACCELERATION itl t' Ed.), reprinted by NATO.
Smith, G. A. and G. Meyer (1979). Total aircraft flight control
system balanced open- and closed-loop with dynamic
trimmaps. 3rd Avionics Conference, Dallas.
. 2 g~| 6a2 ACCELERATION ERROR Smith, G. A. and G. Meyer (1980). Applications of the concept of
OF,~ "~ ~ - dynamic trim control to automatic landing of carrier aircraft.
-.2g NASA TP-1512.
3r er(1) POSITION ERROR Su, R. (1982). On the linear equivalents of nonlinear systems.
Systems & Control Lett., 2, 48.
Su, R., G. Meyer and L. R. Hunt (1981). Transformations of
-3 ~
3~ I"~. V30 nonhomogeneous nonlinear systems..19th A llerton Conference
Communication, Control, and Computing, p. 462.
Wehrend, W. R., Jr (1979). Pilot control through the TAFCOS
i automatic flight control system. NASA TM-81152.
0 100 200 300 400 500
RELATIVE TIME, wc Wehrend, W. R., Jr and G. Meyer (1980). Flight tests of the total
automatic flight cont~ ol system (TAFCOS) concept on a DHC-
FIG. 5. System response--canonic variables. 6 twin Otter aircraft. NASA TP-1513.

You might also like