SNLPart 1

Introduction
inearization
Phase Plane Analysis
Observer design
Control of nonlinear dynamic systems 1
M. Nadri
Dept GEP, Université de Lyon, LAGEPP
Master 3EA, M2 ASI
M. Nadri Control of nonlinear dynamic systems 1 1 / 95

Introduction
inearization
Observer design
Courses Organization
Schedule: 15h CM, 6h TD, 9hTP

Evaluation rules: Practice mark (CP)x 1/3 + final exam mark (CT)x 2/3.
(see MCC voted by the CVU)
Responsable: M. Nadri
Teachers: M. Nadri, madiha.nadri-wolf@univ-lyon1.fr
courses availablity: https://clarolineconnect.univ-lyon1.fr/Non linear
systems 1
Remark
Not everything that is written on the board is on the slides or on the given
documents, mainly the exercises and examples. Be careful, you must write your
notes!

Introduction
inearization
Solution to ordinary di↵erential equations
Observer design
Problem Statement
The 4 steps to control a system
1
figures from Nicolas Marchand, Gipsa/Grenoble
Introduction
inearization
Observer design
Problem Statement
The 4 steps to control a system
Modelling:To get a mathematical representation of the system

Di↵erent kind of model are useful. Often nonlinear
a simple model to analyze the mathematical proprieties and build the
control law
a sharp dynamic model for simulation of the control law and the observer
The state reconstruction: in order to reconstruct the variables needed
for control =) Observer design
Design the controller and test it
Separation principal: Close the loop on the real system
1
figures from Nicolas Marchand, Gipsa/Grenoble
Introduction
inearization
Observer design
Linear systems
A system is called linear if its behavior set satisfies linear superposition

propriety.

Introduction
inearization
Observer design
Linear systems

propriety.
Definition
Superposition principle: 8z1 , z2 2 B and two constants a, b 2 R then
az1 + bz2 2 B.
Control context: Consider two input signals u1 and u2 generating two output
signals y1 and y2 . The system output associated to the input u = a1 u1 + a2u2
is given by y = a1y1 + a2y2 .

Introduction
inearization
Observer design
Linear systems

propriety.
Definition
Superposition principle: 8z1 , z2 2 B and two constants a, b 2 R then
az1 + bz2 2 B.
Control context: Consider two input signals u1 and u2 generating two output
signals y1 and y2 . The system output associated to the input u = a1 u1 + a2u2
is given by y = a1y1 + a2y2 .
Powerful tools founded based on superposition principle make analyzing

the linear systems simple.
Linear systems: can be described by a set of ordinary di↵erential equations
where the closed-form expressions for their solutions are analytically
derivable.
Control and observer design are widely used in the industry.

Introduction
inearization
Observer design
Nonlinear dynamic systems

A nonlinear system is a system that does not satisfy the superposition principle.

Introduction
inearization
Observer design
Nonlinear dynamic systems

A nonlinear system is a system that does not satisfy the superposition principle.
Almost practical systems posses nonlinear dynamics.

Everything is possible:
Many nonlinear phenomena: Multiple isolated equilibrium, Finite escape
time, Limit cycle, Chaos ...
Controllability (resp. observability) are very hard to prove (it is often even
not checked)
Stability may be hard to prove
Local properties 6= global properties (like stability, stabilizability, etc.)
The time behavior is depends upon the initial condition
Frequency analysis is almost impossible
No systematic approach for building a control law: to each problem
corresponds it unique solution
···
But
Sometimes it is possible to describe the operation of physical systems by
linear model around its operating points
Linearized system can provide us an approximate behavior of the nonlinear
system
Mathematical tool: Lyapunov and di↵erential tools
Introduction
inearization
Observer design
Objectives
Give basic tools for modelling of dynamic nonlinear systems

Static nonlinearity handling using first harmonic approximation
Phase plane Analysis and Fundamental pproperties.
Introducing tools such as Lyapunov’s method to analyze the system
stability
Presenting techniques of feedback linearization to control nonlinear
systems.

Introduction
inearization
Observer design
Reference Books
Nonlinear Systems, H. K. Khalil, 3rd edition, Prentice-Hall, 2002

Applied Nonlinear Control, J. J. E. Slotine, and W. Li,
Prentice-Hall, 1991
Nonlinear System Analysis, M. Vidyasagar, 2nd edition,
Prentice-Hall, 1993
Nonlinear Control Systems, A. Isidori, 3rd edition Springer-Verlag,
1995

Introduction
inearization
Observer design
How to build a dynamic model?

Introduction
inearization
Observer design
Continuous time Nonlinear State Model
ẋ1 =f1 (t, x1 , . . . , xn , u1 , . . . , um )

ẋ2 =f2 (t, x1 , . . . , xn , u1 , . . . , um )
....
..
ẋn =fn (t, x1 , . . . , xn , u1 , . . . , um )
ẋi denotes the derivative of xi with respect to the time variable t u1 , u2 , . . . , um

are input variables x1 , x2 , . . . , xn the state variables

Introduction
inearization
Observer design
Continuous time Nonlinear State Model
ẋ1 =f1 (t, x1 , . . . , xn , u1 , . . . , um )

ẋ2 =f2 (t, x1 , . . . , xn , u1 , . . . , um )
....
..
ẋn =fn (t, x1 , . . . , xn , u1 , . . . , um )
ẋi denotes the derivative of xi with respect to the time variable t u1 , u2 , . . . , um

are input variables x1 , x2 , . . . , xn the state variables
2 3 2 3
x1 2 3 f1 (t, x, u)
6 x2 7 u1 6 f2 (t, x, u) 7
6 7 6 u2 77 6 7
6 .. 7 6 6 .. 7
x =66 . 7 , u = 6 . 7 , f (t, x, u) = 6
7 6 . 7
7
6 . 7 .
4 . 5 6 7
4 .. 5 4 .
.. 5
um
xn fn (t, x, u)

Introduction
inearization
Observer design
In matrix form:
ẋ = f (t, x, u), x 2 Rn , u 2 Rm ,
(1)
y = h(t, x, u), y 2 Rp
(1) the state-space model x the state and u the input, y is the output.
Rn = vector space of n-tuples2 of real numbers

state x 2 X , where X is an open subset of Rn or X is a di↵erentiable
manifold 3 of dimension n
input u 2 Rm and output y 2 Rp
X (x0 ; t; u): solution of the dynamical system at time t with initial
condition x0 and input u.
2
An ordered set of n elements. It can be interpreted as a vector, or an n-vector.
3
A manifold is a topological space that is locally Euclidean (i.e., around every point, there is a
neighborhood that is topologically the same as the open unit ball in Rn ), with a globally defined
di↵erential structure
Introduction
inearization
Observer design
Nonlinear systems: state space structures
Linear Time-Invariant (LTI) systems:
f (t, x, u) = Ax + Bu, h(x, u) = Cx + Du
Linear Time-Varying (LTV) systems:
f (t, x, u) = A(t)x + B(t)u, h(x, u, t) = C (t)x + D(t)u
State-affine systems/bilinear:
f (t, x, u) = A(u)x + B(u), h(x) = Cx ( or C (u)x + D(u))
Control-affine systems:
f (t, x, u) = f0 (x) + g (x)u

Introduction
inearization
Observer design
Unforced state equation:

ẋ = f (t, x)
In unforced state equations, input to the system is NOT necessarily zero.

Input can be a function of time: u = (t) a feedback function of the state
u = (x), or both u = (t, x) where is substituted in f (t, x, u)
Autonomous or time-invariant NL systems
ẋ = f (x, u)
function f does not explicitly depend on t.

invariant to shift in time origin, i.e. changing t to ⌧ = t a does not
change f .
The system which is not autonomous is called nonautonomous or
time-varying

Introduction
inearization
Observer design
Remarks
Equilibrium for an autonomous system:

set of real solutions of the equation f (x) = 0.
ẋ = x 2 : isolated equilibrium point
ẋ = sin(x) : infinitely many equilibrium points
ẋ = sin(1/x) : infinitely many equilibrium points in a in a finite region
Linear Systems satisfy the 2 properties:
Homogeneity: f (↵x) = ↵f (x), 8↵ 2 R
Superposition: f (x + y ) = f (x) + f (y ), 8x, y 2 Rnx
4
A system is called forward complete if for every initial condition and every input signal u, the
corresponding solution is defined for all t > 0
Introduction
inearization
Observer design
Remarks
Equilibrium for an autonomous system:

set of real solutions of the equation f (x) = 0.
ẋ = x 2 : isolated equilibrium point
ẋ = sin(x) : infinitely many equilibrium points
ẋ = sin(1/x) : infinitely many equilibrium points in a in a finite region
Linear Systems satisfy the 2 properties:
Homogeneity: f (↵x) = ↵f (x), 8↵ 2 R
Superposition: f (x + y ) = f (x) + f (y ), 8x, y 2 Rnx
Time-varying nonlinear systems
we assume the system to be forward complete4 :
ẋ(t) = f (x(t), u(t), t) ẋ1 = x2

y (t) = h(x(t), u(t), t) ẋ2 = x1 + x12 + sin t
4
A system is called forward complete if for every initial condition and every input signal u, the
corresponding solution is defined for all t > 0
Introduction
inearization
Observer design
Example
m = mass of end weigh

l = length of rod
g = acceleration due to gravity
b = damping coefficient
u = input torque

Introduction
inearization
Observer design
Example

l = length of rod
u = input torque
Equation of motion
ml 2 ✓¨ + b ✓˙ + mglsin(✓) = u

Introduction
inearization
Observer design
Example

l = length of rod
u = input torque
Equation of motion
ml 2 ✓¨ + b ✓˙ + mglsin(✓) = u
Choose state coordinates x1 = ✓, x2 = ✓˙

ẋ1 = x2
g b 1
ẋ2 = sin(x1 ) x 2 + u
l ml 2 ml 2
Introduction
inearization
Observer design
Example: consider di↵erential equations

Introduction
inearization
Observer design
Exercice: Give the model of the system and write it in the state forme (1)

Introduction
inearization
Observer design
Exercice: Give the model of the system and write it in the state forme (1)
- Energy Balance: - Mass Balance:

dE dV (t)
= ⇢in vin Hin ⇢vout H + Q = vin (t) vout (t)
dt dt
E = ⇢VCp H = ⇢VCP (Tj TREF )

Introduction
inearization
Observer design
Example
(
dT (t)
dt
= vVin (t)
(t)
(Tin (t) T (t)) + VQ(t)
(t)⇢C
dV (t)
dt
= vin (t) vout (t)
2 3
 v in (t)
V (t) 6 vout (t) 7
x(t) = , u(t) = 6 7
4 Q(t) 5 , y(t) = T (t)
T (t)
Tin (t)
ẋ1 (t) = f1 (x(t), u(t)) = u1 (t) u2 (t)

ẋ2 (t) = f2 (x(t), u(t)) = ux11 (t)
(t)
(u4 (t) x2 (t)) + u3 (t)
⇢cx1 (t)
y (t) = h(x(t), u(t)) = x2 (t)

Introduction
inearization
Observer design
Definitions and fundamental terminology
1 Recall the axiomatic definition of a K vector space B.

2 Recall the definition of a norm.
3 Recall the definition of a distance between two points in a vector space.
4 Recall the definition of a Cauchy sequence.
5 Recall what a complete space is.
6 Recall the definition of a dot product on a vector space.

Introduction
inearization
Observer design
Recall the definition of a distance between two points of a set.

Introduction
inearization
Observer design

A distance d on a set E is a map d of E ⇥ E in R+ such that:
D’1 8x, y 2 E , d(x, y ) = d(y , x) (symmetry).
D’2 8x, y 2 E , d(x, y ) = 0 , x = y (separation).
D’3 8x, y , z 2 E , d(x, z)  d(x, y ) + d(y , z) (triangular
inequality).
A set endowed with a distance is called a metric space.

Introduction
inearization
Observer design

inequality).
Recall the definition of a norm.

Introduction
inearization
Observer design

inequality).
Recall the definition of a norm.
A norm on a vector space is a mapping from B into R+ denoted k k such
that:
N1 kuk = 0 () u = 0.
N2 8↵ 2 K, 8u 2 B, k↵uk = |↵| kuk.
N3 8u, v 2 B, ku + v k  kuk + kv k
A norm on a vector space defines a distance on this space by:
d(u, v ) = ku v k.

Introduction
inearization
Observer design
Recall the definition of a Cauchy sequence. A sequence (↵n )n2N in a

metric space is said to be a Cauchy sequence if and only if it satisfies the
property:
8✏ 2 R+ , 9p 2 N such that 8n1 > p, 8n2 > p, d(↵n1 , ↵n2 ) < ✏.

Introduction
inearization
Observer design

property:
Recall what a complete space is.

Introduction
inearization
Observer design

property:

A complete space is a metric space in which any Cauchy sequence is
convergent to an element of this space.
Recall the definition of a dot product on a vector space.

Introduction
inearization
Observer design

property:

A complete space is a metric space in which any Cauchy sequence is
convergent to an element of this space.
Recall the definition of a dot product on a vector space.
A scalar product on a vector space E is an application of E ⇥ E in K,
denoted h ; i which verifies the property:
PS1 8↵, 2 K, 8x1 , x2 , y1 , y2 2 E ,
h↵.(x1 +x2 ); .(y1 +y 2) = ↵ hx1 ; y1 ↵ i+↵ hx1 ; y2 i+↵ hx2 ; y1 i+↵ hx2 ; y2 i
PS2 h ; i is symmetric: 8x, y 2 E , hx; y i = hy ; xi.
PS3 8x 2 E , hx; xi 0 and hx; xi = 0 () x = 0. p
We can define a norm from this dot product: 8x 2 E , kxkh ; i = hx; xi.
The notation kxkh i indicates that the norm is induced by the scalar
product.
Introduction
inearization
Observer design
Banach space
It is a vector space B endowed with a norm and which is complete for this
norm.
Example: l p (n) = Rn (set of n numbers from R), p = 1, 2, . . . , 1, p 2 N

equipped with the norm k kp defined by
8 " n #1
>
> X p
>
<8x 2 l (n), p < 1, kxk =
p p
p |xi |
i=1
>
> 1
>
:8x 2 l (n) kxk1 = sup |xi |
i=1,2...,n
Hilbert space
It is a vector space H endowed with a scalar product and (H, h ; i) is complete
for the distance induced by the k·kh i by the scalar product.
A Hilbert space is a Banach space, but not all Banach spaces are Hilbert
spaces.

Introduction
inearization
Observer design
Frame Title
n
X
2 n
For example, l (n) = R with 8x, y 2 R , hx; y i = n
xi yi = x t y is a Hilbert
i=1
space while the others spaces l p (n), p 6= 2 are not Hilbert spaces.
Sequence spaces
The spaces of numerical sequences of summable power p are Banach spaces
and are denoted l p . The space l 2 is, moreover, a space of Hilbert (l 2 is often
used in digital signal processing because F [sampled periodic function] 2 l 2 ).
Functional spaces
Let I ⇢ R be an interval of R.
Examples:
1 I = [a, b] with a, b 2 R,
2 I = [0, +1],
3 I = [ 1, +1]

Introduction
inearization
Observer design
We consider the functions from I in R. Then LpI , also noted Lp (I , R),

designates the vector space (to be proved) of functions of power p integrable
on I :
Z b
1 |f (x)|p dx < 1
a
Z +1
2 |f (x)|p dx < 1
0
Z +1
3 |f (x)|p dx < 1
1
We can define a norm in the following way:
Z b 1
p
1 kf k =
p |f (x)|p dx
a
Z +1
1
p
p
2 kf kp = |f (x)| dx
0
Z +1
1
p
3 kf kp = |f (x)|p dx
1

Introduction
inearization
Observer design
Z
L2I , endowed with the scalar product hf ; g i = f (x)g (x)dx is a Hilbert
I
space. In this space, the norm kf k2 is homogeneous to the energy of the
signal represented by the function f
The set C[a,b] of continuous functions on an interval I = [a, b] endowed

with the norm 8f 2 C[a,b] , kf k = supx2[a,b] |f (x)| is a Banach space.
Indeed, if f 2 C[a,b] , then f is bounded on this interval. It should show, in
addition, that C[a,b] is complete, but it’s a bit long, so we’ll admit it.

Introduction
inearization
Observer design
Lipschitz continuous function
Definition
7 X 0 is Lipschitz continuous when there exists L
A mapping f : X ! 0 such
that
|f (x1 ) f (x2 )|  L |x1 x2 | , 8 (x1 , x2 ) 2 X 2
it is locally Lipschitz continuous when it is Lipschitz continuous on a
neighborhood of every point in X
Lipschitz continuous
+
C 1 ) locally Lipschitz continuous ) C 0
Examples for f : R 7! R
f = x 1/3
f (x) = | sin(x)| ?
f (x) = x 2 ?
p
f (x) = (|x|) ?
Introduction
inearization
Observer design
Example
Let f : R ! R, f (x) = x 1/3 .
f 0 (x) = (1/3)x 2/3

!1ax !0
On the other hand, if f 0 (x) is continuous at a point x0 then f (x) is locally
Lipschitz at the same point because |f 0 (x)| is bounded by a constant k in
a neighborhood of x0 , which implies that f (x) satisfies the Lipschitz
condition with L = k.
Note that the crucial property is that the derivative of f is not bounded
near 0.
Let f : R ! R, f (x) = | sin x|.

Introduction
inearization
Observer design
Example
Let f : R ! R, f (x) = x 1/3 .
f 0 (x) = (1/3)x 2/3

!1ax !0
On the other hand, if f 0 (x) is continuous at a point x0 then f (x) is locally
Lipschitz at the same point because |f 0 (x)| is bounded by a constant k in
a neighborhood of x0 , which implies that f (x) satisfies the Lipschitz
condition with L = k.
Note that the crucial property is that the derivative of f is not bounded
near 0.
Let f : R ! R, f (x) = | sin x|.
Since the cosine is bounded by 1, by the mean value theorem,
| sin x sin y |  |x y | for all x, y 2 R. Therefore, since
ka| |b||  |a b| for all a, b 2 R, we have
|f (x) f (y )|  | sin x sin y |  |x y|
for all x, y 2 R. Hence, f is Lipschitz continuous.

Introduction
inearization
Observer design
Existence and Uniqueness of Solutions
Definition
f (t, x) is piecewise continuous in t on an interval J ⇢ R if for every bounded
subinterval J0 2 J, f is continuous in t for all t 2 J0 , except, possibly, at a finite
number of points where f may have finite-jump discontinuities.
Lemma
Let f (t, x) be piecewise continuous in t and locally Lipschitz in x at x0 , for all
t 2 [t0 , t1 ]. Then, there is > 0 such that the state equation ẋ = f (t, x), with
x(t0 ) = x0 , has a unique solution over [t0 , t0 + ]
Without the local Lipschitz condition, we cannot ensure uniqueness of the

solution.
Example: ẋ = x 1/3 ; x(0) = 0

Introduction
inearization
Observer design
Existence and Uniqueness of Solutions
Definition
f (t, x) is piecewise continuous in t on an interval J ⇢ R if for every bounded
subinterval J0 2 J, f is continuous in t for all t 2 J0 , except, possibly, at a finite
number of points where f may have finite-jump discontinuities.
Lemma
Let f (t, x) be piecewise continuous in t and locally Lipschitz in x at x0 , for all
t 2 [t0 , t1 ]. Then, there is > 0 such that the state equation ẋ = f (t, x), with
x(t0 ) = x0 , has a unique solution over [t0 , t0 + ]
Without the local Lipschitz condition, we cannot ensure uniqueness of the

solution.
Example: ẋ = x 1/3 ; x(0) = 0
x(t) = (2t/3)3/2 and x(t) = 0 as two di↵erent solutions.
The Lemma is a local result because it guarantees existence and
uniqueness of the solution over an interval [t0 , t0 + ], but this interval
might not include a given interval [t0 , t1 ]. Indeed the solution may cease to
exist after some time
Introduction
inearization
Observer design
Global Existence and Uniqueness
Globally Lipschitz
A function f (t, x) is globally Lipschitz in x if
kf (t, x) f (t, y )k  Lkx yk
for all x, y 2 R n with the same Lipschitz constant L.

If f (t, x) and its partial derivatives @fi /@xj are continuous for all x 2 R n , then
f (t, x) is globally Lipschitz in x if and only if the partial derivatives @fi /@xj are
globally bounded, uniformly in t.

Introduction
inearization
Observer design
Global Existence and Uniqueness
Globally Lipschitz
A function f (t, x) is globally Lipschitz in x if
kf (t, x) f (t, y )k  Lkx yk
for all x, y 2 R n with the same Lipschitz constant L.

If f (t, x) and its partial derivatives @fi /@xj are continuous for all x 2 R n , then
f (t, x) is globally Lipschitz in x if and only if the partial derivatives @fi /@xj are
globally bounded, uniformly in t.
Example:
f (x) = x 2 is locally Lipschitz for all x but not globally Lipschitz because
f 0 (x) = 2x is not globally bounded

Introduction
inearization
Observer design
Lemma
Let f (t, x) be piecewise continuous in t and globally Lipschitz in x for all
t 2 [t0 , t1 ]. Then, the state equation ẋ = f (t, x), with x (t0 ) = x0 , has a unique
solution over [t0 , t1 ]
The global Lipschitz condition is satisfied for linear systems of the form
ẋ = A(t)x + g (t)
but it is a restrictive condition for general nonlinear systems

Introduction
inearization
Observer design
Lemma
Let f (t, x) be piecewise continuous in t and locally Lipschitz in x for all t t0
and all x in a domain D ⇢ R n . Let W be a compact subset of D, and suppose
that every solution of
ẋ = f (t, x), x (t0 ) = x0
with x0 2 W lies entirely in W . Then, there is a unique solution that is defined
for all t t0
Example
ẋ = x 3 = f (x)

Introduction
inearization
Observer design
Lemma
Let f (t, x) be piecewise continuous in t and locally Lipschitz in x for all t t0
and all x in a domain D ⇢ R n . Let W be a compact subset of D, and suppose
that every solution of
ẋ = f (t, x), x (t0 ) = x0
with x0 2 W lies entirely in W . Then, there is a unique solution that is defined
for all t t0
Example
ẋ = x 3 = f (x)
f (x) is locally Lipschitz on R, but not globally Lipschitz because

f 0 (x) = 3x 2 is not globally bounded
If for all t, x(t) is positive, the derivative ẋ(t) will be negative and if x(t)
is negative, the derivative ẋ(t) will be positive
starting from any initial condition x(0) = a, the solution cannot leave the
compact set {x 2 R||x|  |a |}
The equation has a unique solution for all t 0
Introduction
inearization
Observer design
Existence and uniqueness of solutions: Particular case
In this lecture, we consider time-invariant NL systems with no input, or with

some fixed feedback u = u(x):
f (x) is a vector field defined on X ẋ = f (x)
A solution of this ODE from an initial condition x0 in X is a function x(t)

defined on an open interval I (x0 ) ⇢ R such that
0 is in I (x0 ) and x(0) = x0
x(t) is C 1 on I (x0 ) with ẋ(t) = f (x(t)) for all t in I (x0 )
I (x0 ) is the maximal interval of definition : there is no other solution
defined on a larger open interval which agrees with x(t) in I (x0 )

Introduction
inearization
Observer design
Sufficient conditions for local existence and uniqueness:
f (x) is continuous on X ) solutions exist for all x0

f (x) is locally Lipschitz on X ) solutions are unique.
Global existence: a locally Lipschitz vector field f is called

Complete when I (x0 ) = R for all x0 in X
forward complete when sup I (x0 ) = +1 for all x0 in X .
Sufficient condition for completeness

f (x) is Lipschitz on X = Rn ) f is complete
X is a compact manifold ) f is complete
Example: non-uniqueness of solutions

f is C 0 but not locally Lipschitz:
p
ẋ = 2 |x|, x(0) = 0, X = R
Try to compute solutions!

Introduction
inearization
Observer design
Check my understanding
What is a globally Lipschitz function ?

Find a continuous function which is not locally Lipschitz ?
Construct a dynamical system which is not complete in positive time

Introduction
inearization
Observer design

Introduction
inearization
Observer design
ẋ = x 2 , x(0) = 1

Introduction
inearization
Observer design
Change of Variables
1
Map: z = T (x), Inverse map: x = T (z)
Definitions
1
a map T (x) is invertible over its donain D, if there is a map T (x) such
that such that x = T (x) 1 for all z 2 T (D)
1
a map T (x) is a di↵eomorphism if T (x) and T (x) are continuously
di↵erentiable.
T (x) is a local di↵eomorphism at x0 if there is a neighborhood N of x0
such that T restricted to N is a di↵eomorphism on N
T (x) is a global di↵eomorphism if it is a di↵eomorphism on R n and
T (Rn ) = Rn
2 @T1 @T1 @T1
3
@x @x2
· · · @xn
6 .1 . . .. 7
6 . . . 7
@T 6 . . . . 7
Jacobian matrix =6
@x 6 .. .. .. .. 77
4 . . . . 5
@Tn @Tn
@x1 @x2
· · · @T
@xn
n

Introduction
inearization
Observer design
Lemma
The continuously di↵erentiable map z = T (x) is a local di↵eomorphism at x0 if
the Jacobian matrix [@T /@x] is nonsingular at x0 . It is a global di↵eomorphism
if and only if [@T /@x] is nonsingular for all x 2 R n and T is proper; that is,
limkx|!1 kT (x)k = 1.

Introduction
inearization
Observer design
Lemma
Example
Negative Resistance Oscillator
 
x2 z2 /"
ẋ = , ż =
x1 "h0 (x1 ) x2 " [ z1 h (z2 )]

Introduction
inearization
Observer design
Lemma
Example
Negative Resistance Oscillator
 
x2 z2 /"
ẋ = , ż =
x1 "h0 (x1 ) x2 " [ z1 h (z2 )]
 
h (x1 ) x2 /" h0 (x1 ) 1/"
z = T (x) = , @T@x
=
x1 1 0
det(frac@T (x)@x) = 1/" is positive for all x
kT (x)k2 = [h (x1 ) + x2 /"]2 + x12 ! 1 as kxk ! 1

Introduction
inearization
Observer design
Remarks: Some coordinates may be more convenient than others

Nonlinear system with X = R2 and x = (x1 , x2 )
ẋ1 = x2 x12
⇣ ⌘
ẋ2 = x1 + 2x1 x2 x12 +u
Consider the di↵eomorphism : X 7! X given by

  
z1 x1 1 z1
z= = (x) = ,x = (z) =
z2 x2 x12 z12 + z2
Same system in new coordinates z :

Introduction
inearization
Observer design
Remarks: Some coordinates may be more convenient than others

Nonlinear system with X = R2 and x = (x1 , x2 )
ẋ1 = x2 x12
⇣ ⌘
ẋ2 = x1 + 2x1 x2 x12 +u
Consider the di↵eomorphism : X 7! X given by

  
z1 x1 1 z1
z= = (x) = ,x = (z) =
z2 x2 x12 z12 + z2
Same system in new coordinates z :
ż1 = z2
ż2 = z1 + u
the system becomes linear when written in the new coordinates!

Introduction
inearization
Observer design
Equilibrium Points
A point x = x ⇤ in the state space is said to be an equilibrium point of

ẋ = f (t, x) if
x (t0 ) = x ⇤ ) x(t) ⌘ x ⇤ , 8t t0
For the autonomous system x = f (x), the equilibrium points are the real
solutions of the equation
f (x) = 0
An equilibrium point could be isolated; that is, there are no continuum of

equilibrium points.

Introduction
inearization
Observer design
A linear system ẋ = Ax can have an isolated equilibrium point at x = 0 (if A is

nonsingular) or a continuum of equilibrium points in the null space of A (if A is
singular)
It cannot have multiple isolated equilibrium points, if xa and xb are two

equilibrium points, then by linearity any point on the form ↵xa + (1 ↵)xb
connecting xa and xb will be an equilibrium point
A nonlinear state equation can have multiple isolated equilibrium points. For
example, the state equation:
ẋ1 = x2 , ẋ2 = a sin x1 bx2

Introduction
inearization
Observer design
A linear system ẋ = Ax can have an isolated equilibrium point at x = 0 (if A is

nonsingular) or a continuum of equilibrium points in the null space of A (if A is
singular)
It cannot have multiple isolated equilibrium points, if xa and xb are two

equilibrium points, then by linearity any point on the form ↵xa + (1 ↵)xb
connecting xa and xb will be an equilibrium point
A nonlinear state equation can have multiple isolated equilibrium points. For
example, the state equation:
ẋ1 = x2 , ẋ2 = a sin x1 bx2
has equilibrium points at (x1 = n⇡, x2 = 0) for n = 0, ±1, ±2, · · ·

Introduction
inearization
Observer design
Linearization
A common engineering practice in analyzing a nonlinear system is to linearize it

about some nominal operating point and analyze the resulting linear model.
What are the limitations of linearization?

Since linearization is an approximation in the neighborhood of an
operating point, it can only predict the ”local” behavior of the nonlinear
system in the vicinity of that point. It cannot predict the ”nonlocal” or
”global” behavior
There are ”essentially nonlinear phenomena” that can take place only in
the presence of nonlinearity.

Introduction
inearization
Observer design
Linearization about a trajectory
Linearization about a trajectory Given: a specific trajectory (x̄(t), ū(t), ȳ (t))

˙
x̄(t) = f (t, x̄(t), ū(t))
ȳ (t) = h(t, x̄(t), ū(t))
Define: error variables
x(t) = x(t) x̄(t), u(t) = u(t) ū(t), y (t) = y (t) ȳ (t)
Compute: linear approximation from Jacobian matrices

 
@f @f
ẋ(t) ⇡ (t, x̄(t), ū(t) x(t) + (t, x̄(t), ū(t) u(t).
@x @u
 
@h @h
y (t) ⇡ (t, x̄(t), ū(t) x(t) + (t, x̄(t), ū(t)] u(t).
@x @u

Introduction
inearization
Observer design
Linearization about an equilibrium trajectory
An equilibrium is a constant trajectory (x̄, ū, ȳ ) and it satisfies 0 = f (t, x̄, ū)
and ȳ = h(t, x̄, ū) for all t.
On the Simple pendulum example

Introduction
inearization
Observer design

Equilibrium trajectory with ū = mg ` sin (x̄1 ) and x̄2 = 0 :
 
0 1 0
ẋ ⇡ g b x+ 1 u
`
cos (x̄1 ) m`2 m`2

Introduction
inearization
Observer design

Equilibrium trajectory with ū = mg ` sin (x̄1 ) and x̄2 = 0 :
 
0 1 0
ẋ ⇡ g b x+ 1 u
`
cos (x̄1 ) m`2 m`2
The LTI system is:

- controllable
- stable when the equilibrium angle x̄1 is in the bottom half of the circle, that is
when |x̄1 | < ⇡2
- unstable when the x̄1 is in the top half of the circle

Introduction
inearization
Observer design
Example1
Consider the following system

⇢
ẋ1 = f1 (x1 , x2 , u) = C1 u x1 k (u x1 )3 x2
ẋ2 = f2 (x1 , x2 , u) = L1 x1
where C > 0, L > 0.

Give the linearized system for the equilibrum point x̄ = [0, 0], ū = 0.

Introduction
inearization
Observer design
Example 2
Let consider system :

⇢
ẋ1 = x2 + ✏x1 x12 + x22
ẋ2 = x1 + ✏x2 x12 + x22
where ✏ 6= 0 and the origin is an equilibrium point.

Give the linearized system:

Introduction
inearization
portrait plan
Observer design
Phase portrait

SNLPart 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SNLPart 1

Uploaded by

Copyright:

Available Formats

Introduction

Control of nonlinear dynamic systems 1

Dept GEP, Université de Lyon, LAGEPP

Master 3EA, M2 ASI

M. Nadri Control of nonlinear dynamic systems 1 1 / 95

Schedule: 15h CM, 6h TD, 9hTP

M. Nadri Control of nonlinear dynamic systems 1 2 / 95

The 4 steps to control a system

The 4 steps to control a system

Modelling:To get a mathematical representation of the system

A system is called linear if its behavior set satisfies linear superposition

M. Nadri Control of nonlinear dynamic systems 1 4 / 95

A system is called linear if its behavior set satisfies linear superposition

M. Nadri Control of nonlinear dynamic systems 1 4 / 95

A system is called linear if its behavior set satisfies linear superposition

Powerful tools founded based on superposition principle make analyzing

M. Nadri Control of nonlinear dynamic systems 1 4 / 95

Nonlinear dynamic systems

M. Nadri Control of nonlinear dynamic systems 1 5 / 95

Nonlinear dynamic systems

Almost practical systems posses nonlinear dynamics.

Give basic tools for modelling of dynamic nonlinear systems

M. Nadri Control of nonlinear dynamic systems 1 6 / 95

Nonlinear Systems, H. K. Khalil, 3rd edition, Prentice-Hall, 2002

M. Nadri Control of nonlinear dynamic systems 1 7 / 95

How to build a dynamic model?

M. Nadri Control of nonlinear dynamic systems 1 8 / 95

Continuous time Nonlinear State Model

ẋ1 =f1 (t, x1 , . . . , xn , u1 , . . . , um )

ẋi denotes the derivative of xi with respect to the time variable t u1 , u2 , . . . , um

M. Nadri Control of nonlinear dynamic systems 1 9 / 95

Continuous time Nonlinear State Model

ẋ1 =f1 (t, x1 , . . . , xn , u1 , . . . , um )

ẋi denotes the derivative of xi with respect to the time variable t u1 , u2 , . . . , um

M. Nadri Control of nonlinear dynamic systems 1 9 / 95

Rn = vector space of n-tuples2 of real numbers

Nonlinear systems: state space structures

Linear Time-Invariant (LTI) systems:

f (t, x, u) = Ax + Bu, h(x, u) = Cx + Du

Linear Time-Varying (LTV) systems:

f (t, x, u) = A(t)x + B(t)u, h(x, u, t) = C (t)x + D(t)u

f (t, x, u) = A(u)x + B(u), h(x) = Cx ( or C (u)x + D(u))

f (t, x, u) = f0 (x) + g (x)u

M. Nadri Control of nonlinear dynamic systems 1 11 / 95

Unforced state equation:

In unforced state equations, input to the system is NOT necessarily zero.

function f does not explicitly depend on t.

M. Nadri Control of nonlinear dynamic systems 1 12 / 95

Equilibrium for an autonomous system:

Equilibrium for an autonomous system:

ẋ(t) = f (x(t), u(t), t) ẋ1 = x2

m = mass of end weigh

M. Nadri Control of nonlinear dynamic systems 1 14 / 95

m = mass of end weigh

M. Nadri Control of nonlinear dynamic systems 1 14 / 95

m = mass of end weigh

Choose state coordinates x1 = ✓, x2 = ✓˙

Example: consider di↵erential equations

M. Nadri Control of nonlinear dynamic systems 1 15 / 95

Example: consider di↵erential equations

M. Nadri Control of nonlinear dynamic systems 1 15 / 95

Example: consider di↵erential equations

- Energy Balance: - Mass Balance: