The Problem of Absolute Stability
Motivation
v0 +
_
u
G(s)
Plant + compensator
u
yR
Actuator
Assume that the actuator is linear
Ky
and assume that th e closed loop system is stable for K [ K min , K max ].
K max
K min
y
Hurwitz sector
8-1
Aizerman conjecture
Assume now that the actuator is nonlinear, for instance, a saturator
( y) u
y
Moreover,assume (t , y), i.e., may change in time.
Would the system be asymptotic
ally stable if (t , y) K min , K max ?
This question was posed by M.A. Aizerman in 1940’s.
Aizerman conjecture : If the closed loop system is asymptotic
ally stable
for all K [0, K * ], it is also asymptotic
ally
stable for all (t , y ) [0, K * ].
8-2
Kalman conjecture
Kalman conjecture : If the closed loop system is asymptotically stable
for all K [0, K * ], it is also stable
for all (t , y ) such that 0
( y ,t )
y
K *.
A answer was first proposed by A.I. Lurie. Popov, Kalman, Yakubovich and
others contributed to the solution. Sometimes this problem is called the Lurie
problem.
N.B : In modern terms, this problem is a problem of robustness:
If we have a nominal feedback ( y ), that is the class
of nonlinearities defined as a sector, where does theasymptotic
al
stability exist ?
8-3
Problem Formulation
Plant x Ax Bu
u Cx
Controller
x Rn , u R p
y Rp
u (t , y)
: R R p R p
Assumptions
( A, B, C )
A
(t, y)
controllab le & observable
Hurwitz
piecewise continuous in t
and satisfies the sector conditions
locally Lipschitz in y
For p 1
y 2 y (t , y) y 2
sector condition
t R
[a, b] local sector condition
y
(, ) global sector condition
8-4
Problem (Continued)
u
y
y
y
Generaliza tion for p 1 (if is decentralized)
We can rewrite the sector condition as
[ (t , y) y][ (t , y) y] 0, t R
[ a, b]
where y or
[, ]
8-5
Problem (Continued)
Consider the decentralized feedback
1 (t , y1 )
(t , y )
2 (t , y p )
Assume each i (t , yi ) satisfies the sector conditions with i , i , ai , bi .
Define K min diag (1 ,, p )
K max diag ( 1 ,, p )
and
{ y R P : ai yi bi }
Then p-dim sector condition is
[ (t , y ) K min y ]T [ (t , y ) K max y ] 0 t R ,
where K max K min 0
y
symmetric positive definite diagonal matrix
8-6
Problem (Continued)
- Generalization for centralized case
Assume L R p p such that (t , y ) satisfies
(t , y ) Ly 2 y 2 t R , y R p
Introduce K min L I
K max L I
Then
[ (t , y ) K min y ]T [ (t , y ) K max y ] (t , y ) Ly 2 2 y 2 0
2
2
where again K max K min 0
Def : A nonlineari ty : R R p R p is called a sector nonlineari ty
if K max , K min R p p such that K max K min is p.d. and symmetric
and [ (t , y ) K min y ]T [ (t , y ) K max y ] 0, t R , y or y R p .
8-7
Problem (Continued)
Notation [ K min , K max ]
or if the inequality is strict, ( K min , K max )
Consider again
x Ax Bu , x R n
y Cx,
u, y R p
u (t , y )
[ K min , K max ]
or
( K min , K max )
(1)
Define : System (1) is absolutely stable in the sector [ K min , K max ]
if eq. point 0 is globally uniformly asymptotic ally stable
for any nonlineari ty from the sector.
8-8
Remarks
Remarks
1)
2)
3)
Absolute stability not another type of stability
Absolute stability gives a measure of robustness
No constructive necessary and sufficient conditions have been
found as yet. The main tool is the Kalman-Yakubovich-Popov Lemma.
8-9
Solution
Approach to the solution
To find out conditions of absolute stability
Find a Lyapunov function good for a continuum of systems – all with
nonlinearities in the sector Two types of Lyapunov functions are
typically used.
V ( x) xT Px
P PT 0
and
V ( x) xT Px 0Cx y ( )d
P PT 0, 0
Here the conditions are less conservative (Popov Criterion).
8-10
Solution (Continued)
In both approaches , the proof of V 0 is done using a technica l result
that allows to complete squares. This is the Kalman - Yakubovich - Popov lemma.
It gives conditions under whi ch a certain stable transfer function Z ( s ) is
strictly positive real, i.e., a stable transfer function Z ( s ) is strictly positive
real if Re Z ( jw) 0, w (, ) or Z ( jw) Z ( jw) 0, w (, )
( just positive real )
If Z ( s) is a ( p p) matrix, the positive realness is defined as follows :
Each Z ij ( s) is analytic in Re s 0
Z * ( s) Z ( s* ) for Re s 0, and
Z T ( s* ) Z ( s) 0 for Re s 0.
Further Z ( s) is strictly positive real if Z ( s ) is positive real
for some 0.
8-11
Kalman – Yakubovich-Popov Lemma
Kalman - Yakubovich - Popov Lemma : Let
Z ( s ) C ( sI A) 1 B D
be an ( p p) transfer function matrix wi th A Hurwitz and ( A, B, C )
observable and controllab le. Then Z ( s ) is strictly positive real if and
only if P PT 0, P R n n , w R p p and L R p n and 0 such that
AT P PA LT L P
PB C T LT w
wT w D DT
8-12
Circle Criterion
Circle Criterion
x Ax Bu
x Rn
y Cx
u, y R p
u (t , y )
At this point, assume A is Hurwitz and [0, K ], that is
T (t , y )( (t , y ) Ky) 0, t , y R p
Choose V ( x) xT Px
Then V xT Px xT Px [ xT AT T BT ]Px xT P[ Ax B ]
xT ( AT P PA) x 2 xT PB
Since 2 T ( Ky) 0, add this to the RHS and obtain an inequality .
Then
V xT ( AT P PA) x 2 xT PB 2( T yT K )
xT ( AT P PA) x 2 xT (C T K PB) 2 T
8-13
Circle Criterion (Continued)
To show when V 0, we use the trick - completion of the squares using
the K - Y - P lemma : Assume P, L & such that
AT P PA LT L P
P PT 0
PB C T K 2LT
Then
V xT Px xT LT Lx 2 2xT LT 2 T
xT Px [ Lx 2 ]T [ Lx 2 ]
xT Px
Now such P, L and do exist as it follows from the K - Y - P lemma
if and only if
Z ( s) KC ( sI A) 1 B I is strictly positive real
8-14
K – Y – P Lemma
K - Y - P lemma : Z ( s ) C ( sI A) 1 B D is strictly positive real iff
A is Hurwitz, ( A, B, C ) controllab le and observable and
AT P PA LT L P
PB C T LT w
wT w D DT
[C ] KC
[ w] 2I
Thus D I
Thus we can show V 0.
8-15
K – Y – P Lemma (Continued)
Lemma : System x Ax Bu
y Cx
u (t , y )
with A Hurwitz and ( A, B, C ) observable and controllab le
and T [ Ky] 0, t , y R p is absolutely stable in the
sector [0, K ] if Z ( s ) I KG( s ) where G ( s ) C ( sI A) 1 B
is strictly positive real.
A graphical illustrati on for the case p 1
Im G( jw)
Re{1 KG( jw)} 0, w
1
Re G( jw)
K
sector [0, K ]
G( jw) is to the right of the vertical line - K1 .
8-16
Generalization
Generaliza tion : Remove condition on A asymptotic stability, sectors ( , ).
0
To eliminate the restriction on A to be Hurwitz
loop transformation (pole shifting)
v0
G(s)
v0
G(s)
K min
(t , y)
K min
y
adding the signal K min y
adding the signal K min y
(t , y)
8-17
Generalization (Continued)
So the scheme does not change. Choose K min so that
( A BK minC )
is Hurwitz. In this new system, the linear part is
GT ( s ) G ( s )[ I K minG ( s )]1
or in the state space representa tion
x ( A BK minC ) x Bu
y Cx
T K min y
Obviously
[ K min y ]T [ K max y ] TT [ K max y K min y K min y ]
TT [ T ( K max K min ) y ] TT [ T Ky] 0 like before
Apply the above lemma : The system is absolutely stable in [ K min , K max ] if
ZT ( s ) I KGT ( s ),
K K max K min
is strictly positive real and
GT ( s ) G ( s )[ I K minG ( s )]1 is asymptotic ally stable.
8-18
Generalization : Case 1
A very convenient representa tion can be given to this condition
if p 1 and K min , K max .
In this case we have
G (s)
GT ( s )
1 G ( s )
G (s)
1 G ( s )
ZT ( s ) 1 ( )
1 G ( s ) 1 G ( s )
To give a graphical interpreta tion, consider three cases seperately :
0, 0, 0
1) 0
Want to find a graphical condition such that GT ( s) is Hurwitz
and ZT ( s) strictly positive real. Begin wit h the latter.
8-19
Generalization : Case 1
1 G ( j )
Re
0, R
1 G ( j )
or
1 / G ( j )
Re
0, R
1
/
G
(
j
)
j Im G( j )
p
1
21
1
Re G( j )
1 G ( j ) is represente d (for point p ) by the line { 1 , p}
1 G ( j ) is represente d (for point p ) by the line { 1 , p}
8-20
Generalization : Case 1
The real part of the ratio of the two complex numbers is positive
when the angle difference is less than / 2
1 2 2
1 G ( j )
Re
0, R
1
G
(
j
)
G ( j ) u jv
Then becomes
(1 u )(1 u ) v 2
0, R
2
2
(1 u ) v
Since the denominato r is positive,
(1 u )(1 u ) v 2 0, R
1
1
(u )(u ) v 2 0, R
8-21
Generalization : Case 1
Im G( j )
G ( j *)
v
1
1
Re G( j )
u
This condition can be given a simple geometric interpreta tion
a :v v :b
v 2 ab
v
a
b
Inequality states that Nyquist plot of G( j ) never enter the disk.
8-22
Generalization : Case 1
Let D( , ) be a disk with the center on the real axis and diameter
1
1
D ( , )
1
1
Then the above condition is met if G ( j ) does not enter D( , ).
Further, GT ( s ) is Hurwitz if the point ( 1 , j 0) is encircled by G ( j )
r times in the counterclo ckwise direction where r is the # of poles of
G ( s ) in the ORHP.
Thus, with p 1, the system is absolutely stable in [ , ] if G ( s ) doesn' t
enter the disk D( , ) and encircles it r times in the counterclo ckwise
direction.
8-23
Generalization : Case 2
2) 0.
Then, GT ( s ) G ( s ),
T
ZT ( s ) G ( s ) 1
We need G ( s ) asym. stable and
Re(1 G ( j )) 0, R
i.e., Re G ( j ) 1
Im G( j )
1
Re G( j )
Thus G( j ) should be right of the vertical line passing through ( 1 , j 0)
8-24
Generalization : Case 3
3) 0
For this case, GT ( s )
1 G ( s)
1 G ( s )
G ( s ) Hurwitz(?)
ZT ( s )
Since
0, the strict positive realness of ZT ( s ) implies
1 / G ( j )
Re
0,
1 / G ( j )
This means that G ( j ) is inside the disk D( , ).
Im G( j )
Thus G ( j ) can' t
encircle ( 1 , j 0)
1
1
Re G( j )
8-25
Summarizing
Summarizin g :
Theorem : Consider a system
x Ax Bu , x R n
y cx
, u, y R
u (t , y ), t , y
( A, B, C ) observable & controllab le
(i) It is absolutely stable in the sector [ , ], 0, if G ( j ) does not
enter D ( , ) and encircles it r times in the counterclo ckwise direction,
where r is the # of G ( s ) poles in ORHP.
(ii) It is absolutely stable in the sector [0, ] if G ( s ) lies to the right of
the vertical passing through (- 1 , j 0) and G ( s ) is Hurwitz.
(iii) It is absolutely stable in the sector [ , ], 0 , if G ( s ) is inside
of D ( , ) and G ( s ) is Hurwitz.
8-26
Summarizing
0 replace G by G and apply (i), (ii)
by
by
If the sector condition is satisfied locally, we have absolute stability
with finite domain.
(1) absolutely stable in [0, 2 ]
(2) absolutely stable in [1 , 1 ]
1
1
1
2
1
1
whi ch is one is bigger?
1
2
whi ch is better for applicatio n?
(3) absolutely stable in [ 2 , 2 ]
8-27
Popov Criterion
Need for Popov' s
1
1
1
2
1
Popov
2
1
1
2 1
Popov Criterion
x Ax Bu , x R n
y Cx
, u, y R
u ( y ) , note that th e feedback is time invariant
Also ( y ) is a decetraliz ed nonlineari ty
8-28
Popov Criterion (Continued)
1 ( y1 )
( y)
p ( y p )
Each i ( yi ) is in t he sect or [0, i ] i 1, , p
i.e.,
0 i ( yi ) yi i yi2 ,
or T ( y )( ( y ) Ky ) 0,
yi (locally or globally)
K diag[ 1 , , p ]
At t his st age, we assume also t hat A is Hurwit z. T his condit ion can
be removed by loop t ransformat ions. As it was point ed out before,
t he main difference bet ween circle & P opov crit erion is t he Lyapunov
funct ion. In t he P opov crit erion, we choose Lyapunov funct ion as follows:
v( x) x T Px 2
y cx
0
T ( ) K d where P 0, P P T and 0
are t o be chosen.
Let F ( y) 2 0y Cx T ( ) Kd 0.
8-29
Popov Criterion (Continued)
For V ( x), we obtain :
d
V ( x) x T Px xT Px F ( y )
dt
F ( y ) dy dx
x T Px xT Px
y dx dt
x T Px xT Px 2 T ( y ) KCx
xT ( AT P PA) x 2 xT PB ( y )
2 T ( y ) KC Ax B ( y )
as for quadratic V ( x)
additional term
Obviously if 0, the additional term is 0 and the two cases are
the same. Adding again
2 T ( Ky ) 0
We obtain
V ( x) xT ( AT P PA) x 2 xT PB 2 T KC Ax B 2 T ( Ky )
xT ( AT P PA) x 2 xT ( PB AT C T K C T K ) T [2 I 2KCB ]
8-30
Popov Criterion (Continued)
Since TKCB is a scalar, V can be rewritten as
V ( x) xT ( AT P PA) x 2 xT ( PB AT C T K C T K )
T [2 I KCB BT C T K ]
Choose small enough so that
2 I KCB BT C T K 0
Now let 2 I KCB BT C T K W TW
Assume that P PT , P R n n , L R p n and 0 such that
AT P PA LT L P
PB C T K AT C T K LTW
8-31
Popov Criterion (Continued)
Then
V xT Px xT LT Lx 2 xT (C T K AT C T K LTW AT C T K C T K ) TW TW
xT Px xT LT Lx TW TW xT LT L TW T Lx
xT Px [ Lx W ]T [ Lx W ] xT Px 0
When P, L & exist? Recall the K - Y - P lemma
Z ( s ) C ( sI A) 1 B D is strictly positive real
iff ( A, B, C ) controllab le & observable , A is Hurwitz
and P, L,W & such that
AT P PA LT L P
PB C T LTW
W T W D DT
So
C T C T K AT C T K
D I KCB
8-32
Popov Criterion (Continued)
Thus
Z ( s ) ( KC KCA)( sI A) 1 B I KCB
I KC ( sI A) 1 B KC[ A( sI A) 1 I ]B
I KG( s ) KC[( sI A)( sI A) 1 A( sI A) 1 ]B
I
I KG( s ) KC[( sI A A)( sI A) 1 ]B
I KG( s ) sKC ( sI A) 1 B
I KG( s ) sKG ( s ) I (1 s ) KG( s )
Thus, the P, L, exist if
Z ( s ) I (1 s ) KG( s ) is strictly positive real.
In addition, the pair ( A, KC KCA) should be observable .
(pair ( A, B) is controllab le by assumption and A is Hurwitz)
Suppose is chosen so that 1 i 0, where i is the eigenvalue of A,
i 1,, n. Then - 1 i and the pair ( A, KC KCA) is observable since
( A, C ) is observable .
8-33
Theorem
Theorem : System
x Ax Bu , x R n , u R p
y Cx
, y Rp
1 ( y1 )
u ( y )
p ( y p )
A Hurwitz
( A, B, C ) controllab le & observable
is absolutely stable in the sector [0, K ) where K diag [ 1 ,, p ]
if 0 with
1
i ( A), i such that
Z ( s ) I (1 s ) KG( s )
is strictly positive real.
8-34
Theorem(Continued)
For p 1, a simple interpreta tion can be given.
Z ( s ) is strictly positive real if
Re[1 (1 j ) KG( j )] 0, R
or Re[ 1 k (1 j )G ( j )] 0
Re[ 1 k (1 j )(Re G j Im G )]
1 k Re G ( j ) Im G ( j ) 0
x
1
k
y
x y 0 y 1 ( x 1k )
Consider the plain (Re G ( j ), Im G ( j ))
Then the above LHS is a line with th e slope
1
intersecti ng
G ( j ) - axis at 1k .
8-35
Theorem(Continued)
Im G ( j )
Graphically
a
1
1k
Re G ( j )
K can be as large as possible
Im G
1k
Re G
Finite sector
Thus with p 1, the system is absolutely stable in [0, K ] if the
Popov plot is to the right of a line intersecti on with the real
axis at 1 k with a slope 1 , 0
8-36
Theorem(Continued)
On the question of choosing so that
2 I KCB BT C T K 0
For p 1, this implies
1 KCB 0
Thus CB is the coefficien t for s n 1 in the numerator of G ( s ).
Indeed,
G ( s ) C ( sI A) 1 B C
Adj( sI A)
B
det( sI A)
b1s n 1 bm
n
s a1s n 1 an
Adj( sI A) [ Is n 1 ( A a1I ) s n 2 ( An 1 a1 An 2 an 1I )]
here ai are the coefficien t of the characteri stic polynomial of A.
Thus CB coefficien t for s n 1.
8-37
Theorem(Continued)
Thus
i) if relative degree of G ( s ) 1, CB 0 and [0, )
ii) if relative degree of G ( s ) 1 but b1 0, [0, )
iii) only when relative degree of G ( s ) 1 and b1 0
1
k | b1 |
i.e., we have to make sure that
1 kb1 0
8-38