Lecture 7: Lyapunov Stability of Nonlinear Systems: Concepts
Lecture 7: Lyapunov Stability of Nonlinear Systems: Concepts
Lecture 7: Lyapunov Stability of Nonlinear Systems: Concepts
Concepts
Stability, asymptotic stability, Lyapunov direct and indirect method
1
7.1.1 Second Lyapunov method
The principle of the second Lyapunov method is to find a function V(x) that represents energy or
generalized energy and satisfies the following conditions.
The equilibrium point, xe=0, is stable Iff:
1. V(x) is continuously differentiable around the fixed point, xe.
V (x) 0 for all x xe 0 and V (0) 0
2. Positive definite V:
dV ( x ) dV (0)
0 at all states x ; 0
3. Negative definite dV/dt: dt dt
The stability of a system trajectories are shown in Fig. 7.2
Stable Unstable
2
2. V(x) is Positive definite V: V ( x ) 0,V (0) 0
3.The derivative, dV/dt, Negative definite: V ( x ) 0,V (0) 0
Note that there are many Lyapunov functions that can satisfy the condition and is therefore not
unique.
Asymptotic Stability
The system is is asymptotically stable at xe=0 under the conditions of Theorem 1, if V(.) if:
(i ) V (0) 0,
(ii ) V ( x ) 0 in D 0,
(iii ) V ( x ) 0 in D 0,
In other words, asymptotic stability is achieved if the conditions of Theorem 1 are strengthened by
requiring dV(x)/dt to be negative definite, rather than negative semi definite.
For the position vector, x(t), Potential energy and kinetic energy are given by:
1 2 1 2
PE x1 x 2
2 2
1 2 1 2
KE x 1 x 2
2 2
A typical energy Lyapunov function, V(x) is in the form of potential or kinetic energy :
3
1 2 1 2
V ( x) x1 x 2 :
2 2
0.5 0 x1
we ca use matrix form: V ( x ) x Ax x1 x2
T
, A is symmetric matrix
0 0.5 x 2
By taking the derivative of the function, the state functions, x’(t), becomes incorporated in the
function so that it has both the function of the function and state trajectory.
V ( x ) x1 x 1 x 2 x 2
x1 x 2 x 2 ( 2 x1 3x2 u(t ) 3x1 x 2 3x 22 x 2 u(t )
V x V V x 1
V ( x ) V(x) * x
x t x1 x2 x 2
x2
x1 x2 x1 x2 2 x1 x2 3x2 x2 u(t )
2
2
1 x 3 x 2 u ( t )
Example
x 1 x 2
Given the system model: x 0 1 x1
x 2 2 x1 ; 1
x 2
2
0 x 2
1 2 1 2
Show that the Lyapunove function is a suitable candidate: V ( x) x1 x 2
2 2
V ( x ) x1 x 1 x 2 x 2 x1 x 2 2 x1 x 2 x1 x 2
Then the derivative: V
f ( x ) x1 ( x 2 ) x 2 ( 2 x1 ) x1 x 2 2 x1 x 2 x1 x 2
t
The dV/dt, is NOT negative definite.
7.1.4 Definiteness
Positive Definite
A function V=f(x) is said to be positive definite in a certain region about the origin if at all points it has
a positive value.
The function V ( x, y ) ( x y ) 2 x 2 y 2
is positive definite if for x=0, or y=0, V(x,y)=0 and for x ≠0, y ≠0, then V(x,y)>0
Positive semi-definite:
A function V=f(x) is said to be positive definite in a certain region about the origin if at all points it has
a positive value except at certain points at which it is zero
The function V ( x, y ) ( x y ) 2 x 2 y 2 2 xy
is positive semi-definite if for x=0, or y=0, V(x,y)=0 and for x ≠0, y ≠0, then V(x,y)>0
The Lyapunov function may be expressed in matrix form as:
1 1 x x y
V ( x, y ) x T Px x y x y x ( x y ) y ( x y x 2 2 xy y 2
1 1 y x y
4
Negative definite:
A function V=f(x) is said to be negative definite in a certain region about the origin if at all points it has
a negative value except at the origin
The function V ( x, y ) x 2 2 y 2 is negative definite.
1 0 x x
V ( x, y ) x T Px x y x y x ( x ) y ( 2 y ) x 2 2 y 2
0 2 y
2 y
Negative semi-definite:
A function V=f(x) is said to be negative definite in a certain region about the origin if at all points it has
a negative value except at certain points at which it is zero
The function V ( x, y ) x 2 y 2 2 xy
is negative semi-definite. When x=0, or y=0, V(x,y)=0 and V(x,y)=0 and for x ≠0, y ≠0, then V(x,y)<0
1 1 x x y
V ( x, y ) x T Px x y x y x ( x y ) y ( x y ) x 2 2 xy y 2
1 1 y x y
Indefinite:
A function V=f(x) is said to be indefinite in a certain region it takes on varying steps
The function V ( x , y ) x y is indefinite
The Fig. 7.2 shows the representation of the various types of equilibrium points and their stability and
the corresponding lyapunov function definiteness.
5
The matrix, Q , can reveal the qualities of the Lyaunov function, v(x). Therefore we obtain the
definiteness of the matrix ,Q.
Symmetric matrices
The quadratic form Q(x1, x2) = 5x21 −10x1x2 +x22 can be represented, for example, by the following 2 × 2
matrices:
5 2 5 3 5 5
A , B , C ,
8 1 7 1 5 1
Matrix C is symmetric: aij = aji.
Theorem 1
Any quadratic form can be represented by symmetric matrix. If a’ijǂ aji we replace them by new
a’ij = a’ji =( aij+aji )/2 ,
This does not change the corresponding quadratic form. Generally, one can find symmetrization
A’ of a matrix A by :
A AT
A' C
2
The Lyapunov’s function can be written using a suitable symmetric matrix, Q,
V ( x ) ax 2 bxy y 2
The matrix Q, is related to the V(x) using the coefficients of V(x) as:
a b x
V ( x) x y 2
y x Qx
T
b 2 c
The matrix, Q, may be used to determine the stability of the system: x Ax
a b
V x T Px, P
Consider a quadratic Lyapunov function candidate b c
where P is a symmetric positive definite matrix.
There are 4 ways to test whether P is p.d.
1. The eigenvalues of P are all Positive.
2. Determinant Test: a>0, ac-b2>0,
3. Pivot Test: a>0, (ac-b2)/a >0,
4. xTPx is positive definite when xe=0.
A symmetric real matrix , Q, is said to be negative definite if x T Qx 0 for all non-zero vector,
xϵRn
Q is positive definite x T Qx 0, xR n \ 0
An nxn symmetric real matrix which is neither positive semidefinite nor negative semidefinite
is called indefinite.
Eigen values
1 2
Consider the matrix A ;
2 1
s 1 2
sI - A ( s 1)( s 3) 4 s 2 4 s 1 0; s1, 2 2 4.472 2.272,6.44
2 s 1
Pivots
1 2
A ;
Consider the matrix 2 1 .
We need to form the matrix to be upper right triangular to get the pivots on the diagonal.
1 2
Subtract (2 *row 1) from row 2, to get A ;
0 3
The pivots are 1,-3 therefore the matrix is not positive definite.
Determinants
If x is an eigenvector of A and λ is an eigenvalue of A, then for x≠0 and Ax = λx. In this case x TAx = λxTx.
If λ> 0, then as xTx> 0 we must have xTAx> 0. For positive definite matrices is xTAx is always positive, for
any non-zero vector x, not just for an eigenvector.
Consider the symmetric matrices A and B :
7
2 1 0
2 1
A ; B 1 2 1
1 3
0 1 2
2 1
For matrix A: |2|>0, 1 3 2 * 3 - 1 5 0 postive definite
2 1 0
2 1
For matrix B: |2|>0, 1 4 - 1 3 0; 1 2 1 6 2 0 4 0 postive definite
2
0 1 2
Where P is a positive definite, real symmetric matrix. Then time derivative of V(x), is:
V ( x ) x T ( Px ) ( x T ) Px
(Ax) T Px x T P ( Ax )
x T AT Px x T PAx
x T ( AT P PA ) x
If V(x) is positive definite, the requirement for asymptotic stability is that V ( x ) is negative.
V ( x ) x T Qx
x T ( AT P PA ) x
1 2 p1 p p1 p 1 0 1 0
0 3 p 0
p2 p
p2 2 3 0 1
p1 2 p p 2 p2 p1 2 p 1 3 p 0 0
3p
3 p2 p 2 p2 3 p2 1 0 0
2 p1 4 p 1 2 p 2 p2 0 0
2p 2p
2 6 p2 1 0 0
8
Solving the simultaneous equations:
1
6 p2 1 p2 0.1667
6
2 p 2 p2 0 p p2 0.167
2 p1 4 p 1 0 p1 0.5 0.333, p1 0.833
0.833 0.167
P ; Det(P) ( 0.833 * 0.1667) - (0.1672 ) 0.167 0
0.167 0.1667
1 2 p1 p p1 p 1 0 1 0
0 3 p 0
p2 p p2 2 3 0 1
p1 2 p p 2 p2 p1 2 p 1 3 p 0 0
3p
3 p2 p 2 p2 3 p2 1 0 0
2 p1 4 p 1 4 p 2 p 2 0 0
4p 2p
2 6 p 2 1 0 0
0.667 0.0832
P ; Det(P) (0.667 * 0.1667) - (0.83 * 0.83) 0.10458 0
0.0832 0.1667
Example 2:
Given system the matrix A and choose the identity matrix, Q=I2, the matrix P can be obtained using:
ATP P A Q 0
0 1 1 0 p p
A ;Q ,P 1
12
8
0 1 p p2
9
s 1
Stability: sI A 12 s 8
s ( s 8) 12 s 2 8s 12 0; s1, 2 4 4 8,0
0 12 p1 p p1 p 0 1 1 0
1 8 p 0
p 2 p p2 12 8 0 1
12 p 12 p 2 12 p p1 8 p 1 0
p 8p 0
1 p 8 p 2 12 p 2 p 8 p 2 0 1
24 p 1 p1 8 p 12 p2 0 0
p 8 p 12 p
1 2 2 p 16 p 2 1 0 0
MATLAB solves the equation AP PA T Q 0 and hence may give a different matrix but is
definiteness is the same.
10
>A=[0 1;-12 -8]; P = [0.4010 -0.5000]
>Q=[1 0;0 1]; [-0.5000 0.8125]
>P=lyap(A,Q)
>eig(P) = 0.0661, 1.1474
11
NB: A Hermitian or symmetric matrix is positive definite iff all its eigenvalues are positive.
Let the Lyapunov function be: V x Px , where P is a positive definite matrix. Given a symmetric
T
Det(A)=ac-b2.
There are four ways to determine if the matrix is positive definite:
i) Eigen value test: λ1<0, λ2<0
ii) Determinant Test: a 0, ac b 2 0
ac b 2
iii) Pivot Test: a 0, 0
a
iv) x T Ax is positive except when x 0
12
f1 f1
f x1 x2 6 2
J
x f 2 f 2 2 6 6 x 22
x1 x2
The Lyapunov function is obtained using the matrix, F:
12 4 12 4
F J JT 2
, | F | 0
4 12 12 x 2 4 12
The matrix F is negative definite over the whole state space. Therefore, the origin is asymptotically
stable, and a Lyapunov function candidate is
6 x1 2 x2
V ( x ) f T ( x ) * f ( x ) [( 6 x1 2 x2 ), ( 2 x1 6 x2 2 x 23 )] 3
2 x1 6 x2 2 x 2
( 6 x1 2 x2 ) 2 ( 2 x1 6 x2 2 x23 ) 2
Since V(x) ∞ as ǁxǁ ∞, then the equilibrium point is globally asymptotically stable.
13