Choice of Lyapunov function for Non-Linear

system
1) Krasovskii’s Theorem
A system described by the dynamic equation:
d
x  f (x)
dt
where x is an n-dimensional vector.
Assuming origin to be an equilibrium point, f(0) = 0
and f(x) is differentiable with respect to each
component of x ,
then Jacobian matrix for the system is defined as:
  f1 f1 f1 
 x 
x2 xn 
 1 
 f 2 f 2 f 2 
 
F ( x)  f ( x)   x1 x2 xn 
x     
 f f n f n 
 n  

 xn xn xn 
• Define F ( x )  F *
( x)  F ( x)
*
where F ( x) is conjugate transpose of F (x).

• If F (x) is negative definite , the equilibrium point
x = 0 is asymptotically stable.
• Then a Lyapunov function for system is
V ( x)  f * ( x ) f ( x)
• And if V ( x)  f ( x) f ( x)   as x  
*

then equilibrium state x = 0 is asymptotically stable

in large.
Proof: Since f ( x)  0 for x  0
and f ( x)  0 for x0
so the V ( x)  f * ( x) f ( x) will be a positive definite
function.
Now we have: d f ( x)   f ( x) d x  F ( x) f ( x)
dt x dt
Next we can find d V (x) :
dt
d
dt
V ( x) 
d *
dt
 
d *  * d
f ( x ) f ( x)   f ( x) f ( x )  f ( x)  f ( x)
 dt  dt
  F ( x ) f ( x ) f ( x )  f * ( x ) F ( x ) f ( x )
*

 f * ( x) F * ( x) f ( x)  f * ( x) F ( x) f ( x )
 
 f * ( x) F * ( x)  F ( x) f ( x)

 f * ( x) F ( x) f ( x)
• if F (x) is a negative definite , then d V (x) will also


be a negative definite function. dt

• Hence, the chosen V(x) is a Lyapunov function.
• Example: Examine the stability of the system
described by following equation by Krasovskii’s
theorem. d
x1   x1
dt
d
x2  x1  x2  x2
3

dt
• Solution: From system equation, we can identify :
f1 ( x)   x1 ; f 2 ( x)  x1  x2  x 3
2
• Obtain the Jacobian matrix as:
 1 0 
F ( x)   2 
 1 ( 1  3 x 2 ) 
 ^
• F ( x) is defined as:
^  1 0   1 1  2 1 
F ( x)   
2   2 
  2 
 1 (1  3x2 )  0 (1  3x2 )  1 (2  6 x2 )
^
• using Slyvester theorem, F ( x) is found to be
negative definite.
• So the equilibrium state at x = 0 is stable.
• Corresponding Lyapunov function is given as:
V ( x)  f ( x) f ( x)  x  ( x1  x2  x )
* 2
1
3 2
2
• It can be seen that V ( x)   as x  
• So the equilibrium state at x = 0 is also
asymptotically stable in the large.
2) Variable Gradient Method
 If a Lyapunov function V(x) which prove stability of a
equilibrium point exists, a unique gradient exists.
 By constraining V(x) to be positive definite and
gradient function to be negative function, The V(x)
can be generated by integrating the gradient.
d
 Let system state space equations be: x  f ( x, t )
dt
 Assuming origin to be an equilibrium point, f(0) = 0.
 Let V(x) be a suitable Lyapunov function, then:
d V . V . V .
V ( x, t )  x1  x2    xn
dt x1 x2 xn
• Above equation can be written in matrix form as:
d  V V V  .
V ( x, t )    x
dt  x1 x2 xn 
• Which can again written as :
d .
V ( x , t )   V  x
T

dt
 V 
 x  
 1
  V1
Where V is the gradient  V   V 
V   x2    2
of V, given by:
     
 V  Vn 
 
 xn 
• V can be obtained from V as a line Integral as:
x1 , x2 ,, xn

V ( x)    V 
T
dx
0
• Upper limit need not imply that V is a vector.
• Integration can be up to an arbitrary point in the
state space.
• Integral can be made independent of path.
• Simple one: Keeping only one of states free while
making other states equal to constant with respect
to the free state.
• So the arbitrary point can be reached along
different coordinate axis of the state space.
 x1 x2

V ( x )   V1  1 ,0, 0  d1   V2  x1 , 2 ,0, ,0  d2  

0 0
xn

  Vn  x1 , x2 , , xn 1 , n  dn
0

• Where Visi the component the xi and

th
i of in V direction
1 , 2 ,  , n
wherex , x ,  , x are dummy variables for
1 2 n
respectively.

• For a scalar function, V(x) to be uniquely

V obtained line integral
of a vector function , then:

Vi
• Following function F(x), formed by taking xmust
j be
symmetric.
  V1 V1 V1 
 x  
x2 xn
 1

 V2 V2

V2 
F   x1 x2 xn 
     
 V Vn Vn 
 n
 
 x1 x2 xn 
• For F to be symmetric, One require following
conditions to be satisfy:

Vi V j
 ; (i  j , i, j  1,2,  , n)
x j xi
• These are n(n-1)/2 equation are called curl
equations
• Problem of determining the Lyapunov function
become equal to the problem of finding a V such
that n- dimensional curl of V equals to zero.
• And also further V and dV/dt must satisfy the
Lyapunov theorems.
• Procedure: Step 1: AssumeV is a general function
of the form: (a x  a x    a x )
 11 1 12 2 1n n 
( a x  a x    a x ) 
V   21 1 22 2 2n n 

  
 
(an1 x1  an 2 x2    ann xn ) 
• Where aij ' s are the unknown parameters that
need to be determined.
Step 2 : From V , determine dV/dt as:
d .

V ( x , t )  V x  T

dt
d V . V . V .
V ( x, t )  x1  x2    xn
dt x1 x2 xn
Step 3: Constrain dV/dt to be negative definite.
Step4: Use the n(n-1)/2 curl equations to be
determine the remaining coefficients of V
Vi V j
 ; (i  j , i, j  1,2,  , n)
x j xi
Step 5: Recheck the dV/dt as the addition of terms must have change
its nature (negative definiteness)
Step 6: Determine V from the line integral given by:
x1 x2

V ( x)   V1  1 ,0,  0 d1   V2  x1 , 2 ,0,  ,0 d2  

0 0
xn

  Vn  x1 , x2 , , xn 1 , n  dn
0

Step 7: Examine region for which conditions on the Lyapunov

function are valid.
Note: In all the attempts to prove stability by second method of
Lyapunov, the failure to prove stability or in the determination of a
suitable Lyapunov function does not mean that system is unstable.
• Example: For system describe by following dynamic
equations:
d d
x1  x2 ; x2   x1  x2
3

dt dt
Find the Lyaponov function so as to proof the
stability of the equilibrium points using variable
gradient method.
Solution: Step-1: Consider gradient of V as:
 ( a11 x1  a12 x2 ) 
V   
( a21 x1  a22 x2 ) 
Step-2: Obtain dV/dt as:
d dx1 dx2
V ( x)  V1  V2
dt dt dt
 (a11 x1  a12 x2 )( x2 )  (a21x1  a22 x2 )(  x1  x2 )
3

 a11 x1 x2  a x  a x  a x x  a21 x1 x2  a x
2
12 2
4
21 1
3
22 2 1
2
22 2

 (a11  a21  a x ) x1 x2  a x  (a12  a22 ) x

2
22 1
4
21 1
2
2

• Step-3: To make dV/dt a negative definite function:

(a11  a21  a22 x12 )  0
a21  0
(a12  a22 )  0
• n=2
• No. of curl equations = n(n-1)/2 = 2(1)/2 =1
• Number of variable to be decided/chosen = 4-1 =3
Choosing a22  3,
from ( a12  3)  0, a12  3, selecting a12  2
a11  a21  3 x12
d
• dV/dt become: V ( x)  a21 x1  x2
4 2

dt
• Gradient of V become:
( a21 x1  3 x13  2 x2 )
V   
 (a21 x1  3 x2 ) 
• Step 4: Remaining one variable can be obtained
from curl equation: V V2
1

x2 x1
 
( a21 x1  3 x1  2 x2 ) 
3
(a21 x1  3 x2 )
x2 x1
  
(a21 x1  3 x1  2 x2 ) 
3
(a21 x1  3 x2 )
x2 x1
• Consider the variable a21 to have a constant part
and a variable part as: a21  a21  a21
(c) (v)

• So,  v
2  a21  a21  x1
(c) (v)
a21
x1
• Relation valid only if:
a21  2 and a21  0
(c ) (v)

• Thus  (2 x1  3 x1  2 x2 ) 
3
V   
 (2 x1  3 x2 ) 
• Step 5: Recheck the dV/dt as the addition of terms
must have change its nature (negative definiteness)
• Step 6: We can determine V by integrating dV/dt along the
special path chosen as above:
x1 x2

V   (21  313 ) d1   (2 x1  32 ) d2

0 0
4 2
3 x 3 x
V  x1 
2 1
 2 x1 x2  2
4 2
4 2
3 x x
V  x12  1  2 x1 x2  x22  2
4 2
4 2
3 x1 x2
  x1  x2  
2

4 2

• Step 7: V(x) is positive definite function in entire region of

state-space.
• Re-Check of dV/dt :
d   2 3 x14 3 x22  dx1
V   x1    2 x1 x2 
dt x1  4 2  dt
  4
3 x1 2
3 x2  dx2
  x1 
2
  2 x1 x2 
x2  4 2  dt
dV
dt
  
 2 x1  3 x13  2 x2 x2   2 x1  3 x2   x13  x2 
 2 x1 x2  3 x1  2 x2  2 x1  3 x2 x1  2 x1 x2  3 x2
3 2 4 3 2

 (2  3 x1  2  3 x1 ) x1 x2   2  3 x  2 x
2
2
4
1

 x  2x
2
2
4
1
• V is a positive definite and dV/dt become negative definite. Thus,
the system is asymptotic stable in large.
 Absolute Stability:
Popov’s Criteria and Circle Criterion
• Reference “Non linear systems” by Hassan K. Khalil, Prentice Hall, 2 nd
edition.
 Tools for the analysis of the feedback systems,
represented by the feedback connection of Figure
given below:
 Many physical systems can be represented by the
feedback connection.
 In a automatic control systems , certain variables are
measured, processed and feedback to manipulate the
control inputs.
 Many system can be remodeled as a feedback systems.
 r 0 Linear system y (t )
u (t )
 G ( s)  C ( sI  A) 1 B


 ( y, t ) Non  Linear y (t )
Element

• Assume that external input is zero and study the behavior of unforced system.
• Use of frequency response of linear system which is based in the classical
control methods like the Nyquist plot and Nyquist criteria.
• The system is said to be absolute stable if it has globally uniformly
asymptotically stable equilibrium point at origin for all non-linearities in a
given sector.
• The circle and Popov criteria give frequency domain sufficient conditions for
absolute stable in the form of strict positive realness of certain transfer
function.
• In SISO system, both criteria can be applied graphically.
 Absolute stability
• Consider the unforced system, represented by:
d
x  Ax  Bu
dt
y  Cx
u   (t , y )

• Where x  R n
; u, y  R p
• (A,B) is controllable and (A, C) is observable.
•  (t , y ) :  0,    R p
 R p
is a memory-less, time-
varying nonlinearity which is piecewise continuous in
t.
• Transfer function of linear system is given by
1
G ( s )  C ( sI  A) B

which is a square strictly proper transfer function.

• Non-linearity element will be required satisfy a
sector condition.
• In case of SISO system, G(s) is SISO transfer
function, p = 1, In that case:
•  (t , y ) :  0,    R  R satisfies a sector condition if
there exits constants,  ,  , a and b ( with    )
and (a  0  b) such that :
y  y (t , y )   y , for all t  0 and y   a, b
2 2

if the above relation holds for all y    ,   , then

one can say that sector condition holds globally.
Then it can be said that  (t , y ) belongs to a sector
 ,  
• Above relation implies the inequality:
 (t , y )  y  (t , y )  y   0, for all t  0 and y   a, b

y This Fig, shows sector

 (t , y ) condition holds globally.
y
y (t ) y
 (t , y )
y
a
b y (t )
• MIMO case: When u and y are p-dimensional
vectors, form of the sector conditions get more
complex.
• Consider the case when the nonlinearity  (t , y )
is decoupled in the sense that  i (t , y ) depends
only on yi , i. e.  1 (t , y1 ) 
 
 (t , y )    
 p (t , y p )

• Each component of  (t , y ) satisfies the sector

condition with constants:
 i ,  i , ai , bi .
• Define
K min  diag(1 ,, p )
K max  diag( 1 ,,  p )
• And
  { y  R P : ai  yi  bi }
• Extending the SISO result to MIMO, one gets:
[ (t , y )  K min y ] [ (t , y )  K max y ]  0 t  R , y  
T

where ( K max  K min )  0

( K max  K min ) is Positive definite symmetric
(diagonal ) matrix
• Above inequality equation will hold for more
general MIMO nonlinearity with different matrices
K min and K max .
• For example: For some matrix L so that

LR p p
such that  (t , y ) satisfies

 (t , y )  Ly 2
  y 2
for all t  0, y    R p

• Where the interior of  is connected and contains

the origin
• Taking:
K min  L  I
K max  L  I

• one gets:
 [ (t , y )  K min y ] [ (t , y )  K max y ]  0
T

2 2
  (t , y )  Ly 2
  y 2
2
 0
K max  K min  0
• where again

• Following definition of a sector non-linearity covers

many non-linearities, both in SISO and MIMO case.
 Definition-1:
A memory-less nonlinearity  ( y , t ) : [ 0, )  R p
 R p
is
said to satisfy sector condition if:
[ (t , y )  K min y ]T [ (t , y )  K max y ]  0, for all t  0, y   or y  R p .
p p
For some real matrices K max , K min  R where K 
( K max  K min ) is a positive definite and where the
interior ofp  is connected and contains the origin.
  R  (.,.)
If , then satisfy the sector condition
 (.,.)
globally , in which case it is said that belong to
a sector [ K min , K max
. ]
If above relation hold with strict inequality, then
 (.,.) is said to belong to the sector ( K min , K max.)
 For all the non-linearities satisfying the sector
condition, the origin x = 0 is an equilibrium point of
the system.
 Aim here is to study the stability of the origin not
for a given non-linearity but for a class of non-
linearities that satisfy a given sector condition.
 If it is shown that origin is asymptotically stable for
all non-linearities in the sector, the system is said to
be absolutely stable.
 This aim was first formulated by A.I. Lurie and hence
sometime called as Lurie’s problem
• Definition-2:
Consider the system given by state equations where
 (.,.) satisfies a sector condition as per definition-1. The
system is absolutely stable if the origin is globally
asymptotically uniformly stable for any non-linearity in the
given sector. It is absolutely stable with a finite domain if
the origin is uniformly asymptotically stable.

Remarks
1) Absolute stability  not another type of stability
2) Absolute stability gives a measure of robustness
3) The main tool is the Kalman-Yakubovich-Popov Lemma
 Solution to Lurie’s problem
• Approach to the solution
To find out conditions of absolute stability
Þ Two types of Lyapunov functions are typically used:
V ( x)  x Px
T
PP 0 T

and
y
V ( x)  x Px     T ( ) K d
T
P  P T  0,   0
0

ÞFirst one is simple quadratic form

ÞSecond one is known as Lurie type Lyapunov function. In this
case the nonlinearity is time invariant and satisfy some
conditions to ensure well defined integral and non-negative
 In both cases, starting is done with derivation
(dV/dt) with P undermined.
 Then determine the conditions under which there
exists P such that dV/dt along trajectories of the
system is negative definite for all nonlinearities that
satisfy a given sector condition.
 In the Circle criteria, these condition takes the form
of a frequency-domain conditions on a certain
function.
 In the Popov’s criteria, the search for P is replaced
by an optimization problem whose solution is
sought numerically.
• Definition-3
• A p X p proper rational transfer function matrix Z(s)
is called positive real if:
a) All elements of Z(s) are analytic for Re[s]  0
b) Any pure imaginary pole of any element of Z(s) is
a simple pole and the associated residue matrix
of Z(s) is positive definite Hermitian and,
c) for all real , for which j is not a pole of
any element of Z(s), the matrix Z ( j )  Z T ( j )
is positive semi-definite.
The transfer function matrix Z(s) is called strictly
positive real if Z ( s   ) is positive real for some   0.
Kalman – Yakubovich-Popov (KYP) Lemma
Kalman-Yakubovich-Popov Lemma also known as
positive real lemma.
It is stated as:
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  P T  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
 Circle Criterion
• Consider the system
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  x T 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 RHS and obtain inequality :
Then V  xT ( AT P  PA ) x  2 xT PB  2( T  y T K )
 xT ( AT P  PA ) x  2 xT (C T K  PB)  2 T
 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 2 xT 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
• Comparing with K-L-Y Lemma, one gets:
[C ]  KC
[ w]  2I
Thus D  I
• Thus we can show that dV/dt < 0 (negative definite)
• Writing the result in form of a Lemma:
System x  Ax  Bu
y  Cx
u   (t , y )
with A Hurwitz and ( A, B, C ) observable and controllable
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.
 This means
Re{1  KG ( jw)}  0, w
A graphical illustrati on for the case p  1  SISO
Im G ( jw)

1 Re G ( jw)
K

sector [0, K ]

G ( jw) is to the right of the vertical line - K1 .

 Multivariable Circle criteria
d
• Consider the system as given by: x  Ax  Bu
dt
y  Cx
u   (t , y )
• where (A,B) is controllable ,(A,C) is observable and non-
 (t , y )
linearity satisfy the sector condition globally then the
system is absolutely stable if 1
GT ( s )  G ( s )[ I  K min G ( s )]

• Is Hurwitz and
Z T ( s )   I  K max G ( s ) I  K min G ( s )
1

• Is strictly positive real.

 R p If the sector condition is satisfied Zonly
T (s )
on
a set GT (s) then the conditions given on and ensures
that the system is absolutely stable with a finite domain.
Popov’s Criteria
• Consider the system:
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.

  1 ( y1 ) 
 
 ( y)    
 p ( y p )
 
Each  i ( yi ) is in the sector [0,  i ] i  1,  , p
i.e., 0   i ( yi ) yi   i yi2 , yi (locally or globally)
or  T ( y )( ( y )  Ky )  0, K  diag [ 1 ,  ,  p ]
 At this stage, we assume also that A is Hurwitz.
This condition can be removed by loop transfor mations.
In the Popov criterion, we choose Lyapunov function as follows :
y  cx
V ( x)  x Px  2 
T
 T ( ) K d where P  0, P  P T and   0
0

are to be chosen.
Let F ( y )  2 0y Cx T ( ) Kd  0.
For V ( x), we obtain :
d
V ( x)  x Px  x Px  F ( y )
 T T

dt
F ( y ) dy dx
 x Px  x Px 
T T

y dx dt
 x T Px  xT Px  2 T ( y ) KCx
 xT ( AT P  PA ) x  2 x T PB ( y ) as for quadratic V ( x)
 2 T ( y ) KC  Ax  B ( y ) 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  2KCB]
Since  TKCB is a scalar, V can be rewritten as
V ( x )  x T ( AT P  PA ) x  2 x T ( PB  AT C T K  C T K )
 T [ 2 I  KCB  B T C T K ]
Choose  small enough so that
2 I  KCB  B T C T K  0
Now let 2 I  KCB  B T C T K  W T W
Assume that  P  P T , 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  LT W
Then
V  x T Px  xT LT Lx  2 x T (C T K  AT C T K  LT W  AT C T K  C T K )  TW T W
 x T Px  xT LT Lx  T W T W  x T LT L   T W T Lx
 x T 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 ) controllable & 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
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 controllable 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.
 Multivariable Popov 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 ) controllable & observable
is absolutely stable in the sector [0, K ) where K  diag[ 1 ,,  p ]
1
if   0 with   i ( A), i such that

Z ( s )  I  (1  s ) KG ( s )
is strictly positive real.
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 y
1
k  x   y  0  y  1
 ( x  k)
1

Consider the plane (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 .
  Im G ( j ) Graphically

1
 1k 
a
Re G ( j )

K can be as large as possible

Thus with p  1, the system is absolutely stable in [0, K ] if the

Popov plot is to the right of a line intersection with the real
axis at  1
k with a slope 1
 ,  0
Generalization
Generalization : Remove condition on A asymptotic stability, sectors ( ,  ).

To eliminate the restriction on A to be Hurwitz

loop transformation (pole shifting)
v0   y
v0 G(s)
G(s)  
 
K min
adding the signal  K min y

 (t , y ) K min adding the signal  K min y


  (t , y )
So the scheme does not change. Choose K min so that
( A  BK min C )
is Hurwitz. In this new system, the linear part is
GT ( s )  G ( s )[ I  K min G ( s )]1
or in the state space representa tion :
 x  ( A  BK min C ) 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 asymptotically stable.

Thus we can conclude with writing result in form of a

theorem which is refer as multivariable circle criteria: