Professional Documents
Culture Documents
Absfracf-Necessary and sdicient conditionsfarthe "decou- proof of sufficiency, and they did not consider the ques-
pling" of an m-input, m-output time-invariant linear system using tion of necessity.
state variable feedback are determined.Given a system which satis-
fies these conditions, i.e., which can be decoupled by state variable
In this paper, a necessary and sufficient condition for
feedback,theclass of all feedbackmatriceswhichdecouplethe decoupling will be given; a characterization of the class
system is characterized. The characterization of @ is used to deter- of feedback matrices which decouple a system u-ill be
mine the number of closed-loop poles which canbe specified for the determined; the number of closed-loop poles which can
decoupled system and to develop a synthesis technique for the real- be specifiedwhile decoupling will also be determined;
ization of desired closed-loop pole configurations. Transfer matrix
and a synthesis procedure for obtaining desired closed-
consequences of decoupling are examined and practical implications
discussed through numerical examples. loop pole configurations will be developed. In line n-ith
these objectives, the remainder of this paper is divided
I. INTRODIJCTION into thefollowing sections:
11. Definitions
H ED E V E L O P M E N T of techniquesforthe
111. Main Theorem
design of multivariable control systems is of con-
11'. Class of Decoupling 3Iatrices
siderablepracticalimportance. A particularde-
sign approach involves the use of feedback to achieve
v . A Synthesis Procedure
VI. Decoupling by Output Feedback
closed-loop controlsystemstability.Inconjunction
VII. A Practical Example.
withthisapproach,it isoften of interesttoknow
whether or not it is possible to have inputs control out- In Section I1 precise definitions of state variable feed-
putsindependently, Le., a single input influences a back and decoupling are given. Then the basic necessary
single output. This is, in heuristic terms, the problem of and sufficientcondition for decouplingisprovedin
decoupling. Section 111. Using the main theorem, a descriptionof all
T h e problem of decoupling a time-invariantlinear thedecouplingmatrices is presentedinSection 117.
system using state variable feedback and the relation of Next, in Section V the questionsof synthesis andclosed-
thisproblem to control system design have been dis- loop pole placement are examined. In Section VI state
cussedbyseveralauthors. [11--[31 hiorgan[']considered variable feedback is replaced by output feedback and
thequestion of decouplingforsystemsn-hosestate the relevant theory developed. The practical potential
equations had a somewhat special form. His main re- of themethods is indicated in the discussion of a
s u l t , which is a special case of the main theorem of this VjSTOL stability augmentation systemin Section VII.
paper,wasthe following: thetime-invariantlinear
system 11. DEFIXITIONS
x=Ax+Bu Consider the time-invariant linear system
X=Ax+Bu
y = cx
y = cx (1)
canbedecoupled if thematrix C B isnonsingular.
where x is an n vector called the state, u an m vector
Rekasius[*] extended Morgan's result 'and outlined an
called the control (or input), y an m vector called the
essentiallytrial-and-errorprocedureforspecifying a
output, and A , B, and C are n X n , n X m , and YIZ X n
certain number of the system's poles while decoupling
matrices, respectively. I t is assumed that n z s n . If F is
the system. Xeither IIorgan nor Rekasius gave a clear
an m X n matrix and G a nonsingular n z X m matrix, then
the substitution of
First partof manuscript received Sovember 28, 1966; secondpart
of manuscript received March 29, 196;. First part revised April 19, u = Fx 4- Go (2)
1967; first and second parts combined and revised J u l y 14, 1967.
This research was supported in part by the Air Force Office of sci- whererepresentsthenew m vectorcontrol(Fig. I),
entific Research under AFOSR Grant 693-67 and in part by N.-2S.I.
P. L. Falb is with the Div. of AppliedMathematics, Brown into (1) shall be called linear stafe variablefeedback.
University,Providence,R.
Center, Cambridge, Mass.
I., and KASd ElectronicsResearch -
Let d l , d?, . . , d m be given by
IV. A. ITolovich is with NASA ElectronicsResearchCenter,
Cambridge, Mass.
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
652 AUTOMATIC
TRANSACTIONS
ON IEEE CONTROL, DECEMBER 1967
a; = tz - 1 if CiAjB = 0 for a l l j
I:=O
x-here Cidenotes the ithrow of C. Then, a simple calcu-
lation shows that
= t r ( L i ( F GIG'),
,
. , ill. (11)
for i = I , . . . , m.. Application of the state variable feed- Note that this is a precise definition t h a t does not
back (2) and repeated differentiation togetherwith (4) involvevaguestatementsaboutinputscontrolling
yield the relations outputs independently.
Yi =C ~=
X C;(A + BF)OX 111. AIAIS THEOREM
y
i X C i ( A + BF)x
= C~A= IVith the definitions of Section 11, it is no\v possible
to state and prove a theorem which gives a necessary
and sufficient condition for decoupling.
(12)
C,Adm+l
]
+
C i [ ( A BF)di]BG
0 J
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
FALB AND WOLOVICH: DECOUPLING IN MULTIVARIABLE CONTROL SYSTEMS 653
where
Qi(F) = [ Ci(A
Ci(A
+ BF)n-’B
Ci(A + BF)n-2B
\
+ BF)d’B
1, i
of
in the proof of the theorem is thefollowing observation. Q i ( F ) . Let Pi(F) be the n X n matrix given by
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
634 DECEMBER
AUTOMATIC
CONTROL,
TRANSACTIONS
ON IEEE 1967
Since P'(F) is nonsingular, it follon-s that the rank If the pairF,G decouples (l),then there is adiagonal
of Pi(F)Qi(F)is the same as the rank of Qi(F). Kote matrix A such that G=AB*-l.
also t h a t Proof: If F, G decoupless (l), then Q i ( F ) is given by
(37) and
L i { F ,G } = Pi(F)Qi(F)G (3 2)
.
I xis+ I
.I "y 1
1
l xi
I
l
.. .I
I
1
0 1 ; Xi#(), (40)
Theorem 2 I t follo\vs
B*G=diag
that [ X 1 , . . , X.],theand cod-
If the pair F, G decouples ( l ) , then the rank of Q i ( F ) 1aV is established.
is one for all i ; conversely, if the rank of Qi(F)is one for
all i and if B* is nonsingular,thenthepair F, E*-' Corollary 2
decouples the system (1). If the pair F,G decouples (l),then there is a diagonal
Proof: Supposefirstthat F, G decouple (1). Thenmatrix such that ~
tr ( L ~ { FG, ] Q ) = tr ( L ~ ( FG)Q'>
, z o (34) FB = B*-l{ ,,A** - A*}B
(41)
forall i where a is the m X l z matrixgivenbywhere A** and A* are given by
Since Parbitrary,
p
is column
ith
the
=[ I
I
I
I
...j
1
"(n-1)
I
of L i ( F , G } is a
. (35)
A** = [ ] CIA''
C,AL
, A* = A**A.
(42)
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
FALB iWD WOLOVICH:
MULTIVARIABLE
DECOUPLING
IN CONTROL SYSTEMS 655
A=[::
1 1 0
:I, .-[d b], varyingthe Mk. Inthislight,considerthe
example.
Example 2
Let
following
1 0 0
= [o 0 11. - 0 1 0 0 0 0 - -0 1
Then 0 0 1 0 0 0 1 0
B* = ['IB
C2AB
] [
=
-1
'1.
1
A =
0
0
0-1
0 0
0
0 1
0 0
0 '
B=
0
0
0
0
0 0 0 0 0 1 0 0
Thus, B* is nonsingularandthesystemcanbede-
-0 0 0 0 0 0- ,1 0
coupled. The set @ of all F which decouple the system
(45) can now beobtainedbydetermining all 2x3
matrices F such that rank [ Q i ( F ) ]= 1. In this example,
this implies that the elements of SP must be of the form
V. A STXTHESIS
PROCEDURE
Theorem 2 does provide a procedure for determining N O m
a, the class of all feedback matrices F which decouple
(1). However, the direct application of the condition,
rank [ Q i ( F ) ]= 1 forall i, results only in constraints
B* = A*=[ [y ,]A
0 1 0 0 0 0
0 0 0 0 0 0
being placed on certain of the nzn parameters of F.W h a t Since B* is nonsingular, the system can be decoupled.
is still required is a procedure for specifying closed-loop Setting, for example,
system poles while simultaneously decoupling (1) using
an appropriate FE@. Inthislight, a synthesispro-
cedure will now be presented for directly obtaining a
feedback matrix F e @ whose parameters are so deter-
mined as to yield a desired closed-loop pole structure. one obtains, using (29), the decoupled system
In particular, suppose that M k , R =0, 1, . . . , 6 are
given m X m matrices. Then the choice
6
F = B*-'[ M&Ak - A*]
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
656 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, DECEMBER 1967
Lemma Example 3
Let K be the (m+ x?=,X n matrix given by
di) Let
"1
0 1 0 0 1
A=
1 0 '
O o0 10 O 0 I '
0 B=
L o o o IJ 0 1
K=
rl 0
c ' = J 0 0
0 0
Lo 1
Then rank [K] =m + x;"=
d i and hence m di 4n. + Then
Proof: Let ki denote the ith row of K, and let ri be
arbitrary scalars such that B* =
C1A2B
[c2B =I ]
2 1
riki = 0 (57) and m+ d i = 4 =n. Thus, all the closed-looppoles
can be specified by using the synthesis procedure.
where
m
v =m + 1
di.
1
0
I
Now let p denote the number of closed-looppoles Example 4
which can bespecified while decoupling, and letf denote Let
the number of free parameters (entries) in a decoupling
0 1 0
1
matrix F [for example, f = 3 in (47)]. Then the lemma 0 0
and (51) combine to give A= B=[l 0
m [ 21 31 011 , 0 1
n z f ~ d i I p I n 1 1 0
1
= co 0 11.
Then
Moreover, if m f cy
di= n, then all n of the closed-
loop poles can be arbitrarily positioned while simultane-
ouslydecouplingthe
system. Also, if f = m + di,
B* = [zz] = I.
then (51) gives direct physical significance to the free Thus, B* is nonsingular and the system (64) can be de-
parametersin F. If f >m+ di (or n ) , then i t may coupled. I t can be shown that the elements
of @ must be
be
possible to
specify
morethan .m+ di
of
the such that
closed-loop poles. In this situation, it is often advanta-
geous to calculate C(s1- A-BF)-lBB*-l with f entries
in F remainingarbitrary.The following examples
illustratetheseideasandsome
volved in their application.
of the difficultiesin- so t h a t f = n = 3 > 2 = m+
loop transfer matrix is given by
xy
di. Moreover, the closed-
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
FALB AND WOLOVICH:
DECOUPLING
MULTIVARIABLE
IN CONTROL
SYSTEMS 657
-
-
( S S 1) (
(s-jz3-1)
s;j2r 1) 0
S2-(fi2f3)S-((fi~f2)
[~'-((f12+3)S-((fll+2)]
1 (65)
so t h a t all of the closed-loop poles can be specified by Since det(sI-A-BHC) = (s- 1-hz2)(s3-h11), output
application of Theorem 2. However,notethatappli- feedback will not be adequate to stabilize the system
cation of thesynthesisprocedure in this casewould although state variable feedback provides a higher de-
allow one to specify only two of the three closed-loop gree of flexibility (63).
poles.
Example 6
I?I. DECOCPLING FEEDBACK
BY OUTPUT
Consider the system described by (64). I t has been
Since output feedback is only a special case of state shown in ( 6 7 ) that state variable feedback can be used
variable feedback, i.e., to decouple the system while simultaneously specifying
u=Hy+Go=HCx+Go (68) all three closed-looppoles. Application of Theorem 2
and (39) imply that any2 x 2 matrix H of the form
with H C replacing F , i t follows immediately that (1)
can be decoupled using output feedback if and only if (74)
1) B* is nonsingular and 2) there is an m X m matrix H
such that rank [Q'(HC)] = 1 for i = 1, . . . , m. These
conditions provide a suitable test for whether or not a will define anoutputfeedback n-hichdecouplesthis
system can be decoupled using output feedback. system. From ( 7 4 ) it follows t h a t
det(sl-A-BHC)=(s-l-h~~)(s2-(h~+3)s-(hll+2))
Example 5
and hence that the system can be stabilized using out-
Let
put feedback (e.g., hZ2=- 1, kll= - 5 ) , although the
poles are not completely arbitrary.
Example 7
1 0 0 Let
c = [o 0 11-
Then
B* = [-1 '1
1
is nonsingular so that the systemdefined by (69) can be
decoupled. However, i t is not possible to decouple this
system using output feedback.T o see this, observe that Then
c' = L -1 0
m
m + ai = 2 <3 = n.
1
a n d t h a t H C m u s tbe of the form
I t can be shown, using Theorem 2 , that any decoupling
F is of the form
0 fl2
Equations (71) and ( 7 2 ) lead to the contradictory re-
quirement that flz = 0 and flz = - 1. This example illus-
= [, f2Z -1 - f22
f131
(77)
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
658 AUTOMATIC
TRAiiSACTIONS
ON IEEE CONTROL, DECEMBER 1967
- 1212
(s - l)(s -
- l)(s - h21
0
1221
- 1)
- 1) 1 (79) and is nonsingular since it is assumedZcsl?f~nfXcr;
that zO.
Therefore
so that output and state variable feedback are equiva- 3
1
0 0 0 0
T 1
zcs ?n2' WZ20
0 0 0 1 0 0 0
0
where
1 0 0 0 0
24 =incremental
longitudinal ( x ) velocity change,
8 =incremental pitch angle, 0 0 0 1 0 0
O I
e
= pitch rate,
If G is now set equal to B*-l, the closed-loop transfer
w =incremental vertical (2) velocity change, matrix is
Ax = incremental position error,
AZ =incremental altitude error, C(d-A -BF)-'BB*-'
6, =incremental collective fan input,
= incremental nose fan input,
6, =incremental fan stagger input.
0,
0,
(S'-17t11S-?~Z10)(~-m31S-~30),
0, (s2-nzllS-nzlo)(s2-~~p'S-m20)
(85)
(s2-~~z11s-~~~10)(s2-~~z21--1n20)(s2-m31s-?~.30)
0,
:I -
If the m i i are suitably chosen, then, in effect, the pilot
The output matrix C is thus defined as
will be faced with the task of controlling three highly
0 1 0 0 0 0 stable second-order systems. This example serves only
to indicate a potential practical area of application for
c = 0 0 0 0 1 0 .
the ideas presented in this paper.
[ o o o o o l l (81) Theaboveexamplesillustratethetechniquesde-
velopedfor
synthesizing decoupling
controllers
for
1 Similar to the XV-5.4. multivariable systems.
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.
F.4LB AND WOLOVICH:
DECOUPLING IN MULTIVARIABLE
CONTROL
SYSTEMS 659
Authorized licensed use limited to: PAKISTAN INST OF ENGINEERING AND APPLIED SCIENCES. Downloaded on March 14,2021 at 06:17:58 UTC from IEEE Xplore. Restrictions apply.