Nonlinear Systems

Introduction
 Nonlinear Systems 1

✔ In general, analysis and control design is performed on the mathematical model of

the system.

✔ Laws of natural/engineering sciences and simplifications are used to obtain the

mathematical model.
➙ Expectations/requirements from the system are also considered.
✔ Analysis: Given the inputs, initial conditions, and system model/parameters or
changes/variations in those determine the response of the system.

✔ Design: Determine inputs/controllers which will lead to the desired

behavior/response.
 Nonlinear Systems 2

✔ The dynamic response of a system may include responses to

➙ Changes in the external effects/inputs
➙ Changes in the internal variables/states
➙ Changes in the system parameters
✔ From engineering perspective the main reason for trying to understand the
dynamics of a system is the desire to control it (in order to achieve desired
response/performance).
 Nonlinear Systems 3

✔ In previous courses, you considered systems which are linear or linearized about
an operating point

✔ Results obtained for linear systems hold globally

➙ Stability
➙ Controllability
➙ Observability
✔ For nonlinear systems
➙ The above mentioned properties might hold only locally
➙ There might be multiple isolated equilibrium points
➙ Initial conditions are important - different initial conditions may lead to different
behavior

➙ Linearization about different operating points may lead to different linear

equivalent linear systems

✔ Real physical systems are in general nonlinear

 Nonlinear Systems 4

In this course

✔ It is assumed that the students are familiar with the fundamental concepts in
control systems and linear systems

✔ We will improve our knowledge/understanding about linear systems

✔ We will learn new knowledge/skills on nonlinear systems
➙ Some fundamental analysis techniques
➙ Some controller design techniques (mostly in a follow-up course)
✔ We will improve our background in systems theory
✔ We will get ready to read, understand, and follow the literature on nonlinear
systems theory and control

✔ We will get ready to do research on nonlinear systems theory and control

 Nonlinear Systems 5

Comparison of Linear and Nonlinear Systems

Let y1 = h(u1 ) and y2 = h(u2 )

In linear systems, for all inputs u1 , u2 and for all coefficients α1 , α2

h(α1 u1 + α2 u2 ) = α1 y1 + α2 y2
is satisfied.

✔ In linear systems the superposition principle is satisfied.

✔ Nonlinear systems do not satisfy the superposition principle.
✔ If a system is linear dynamical system in state space form, the superposition
principle is satisfied for all state-input pairs {u1 , x1 } and {u2 , x2 }
 Nonlinear Systems 6

Example:
y = 5u → linear
y = u2 → nonlinear,
y = sin(u) → nonlinear
y = u1 u2 → nonlinear (and multi-input system u = [u1 , u2 ]⊤ )
 Nonlinear Systems 7

Dynamic System Model (State Space Model)




 ẋ1 = f1 (t, x1 , ..., xn , u1 , ..., up )

= f2 (t, x1 , ..., xn , u1 , ..., up )

 ẋ
2
✔ State equations :


 ...


 ẋ
n = fn (t, x1 , ..., xn , u1 , ..., up )
✔ ẋi is the derivative of state xi with respect to (wrt) time t
✔ x1 , ..., xn are the state variables of the system (its ”memory” about the past
inputs)

✔ u1 , ..., up are the input variables of the system (applied to the system)



 y1 = h1 (t, x1 , ..., xn , u1 , ..., up )

= h2 (t, x1 , ..., xn , u1 , ..., up )

 y
2
✔ Output equations:


 ...


 y
q = h (t, x , ..., x , u , ..., u )
n 1 n 1 p
 Nonlinear Systems 8

     
x u y
 1   1   1 
 .   .   .
✔ Defining x =  ..  ∈ Rn , u =  ..  ∈ Rp , y =  ..  ∈ Rq ,

     
xn up yq
   
f1 (t, x, u) h1 (t, x, u)
   
.. n ..
f (t, x, u) =  ∈ R , h(t, x, u) =  ∈ Rq
   
.   .
   
fn (t, x, u) hq (t, x, u)

➙ The system can bewritten in the form

 ẋ = f (t, x, u) (state equation)
State space form:
 y = h(t, x, u) (output equation)

✔ x = state vector, u = input vector, y = output vector

 Nonlinear Systems 9

Special Cases

 ẋ = A(t)x + B(t)u
✔ Linear time-varying (LTV) system:
 y = C(t)x + D(t)u

 ẋ = Ax + Bu
✔ Linear time-invariant (LTI) system:
 y = Cx + Du
n
✔ Unforced state equation: ẋ = f (t, x)

➙ Does not imply that the system is without input. It could be that there is a
controller of the form u = γ(t) or u = γ(x) or u = γ(t, x) resulting in the
above form.
n
✔ Autonomous system: ẋ = f (x)
 Nonlinear Systems 10


 ẋ = f (x, u)
✔ Time-invariant system:
 y = h(x, u)

➙ In time-invariant systems, if the input is shifted and applied at time t0 + a instead

of time t0 , then the response of the system is shifted with the same amount of time
a (provided that the initial conditions are the same).
➙ The response of time invariant systems is independent of the time at which the
input is applied, it depends only on the input itself (and the initial conditions).
 Nonlinear Systems 11

Equilibrium Points

✔ A point x = x∗ in the state space is called an equilibrium point of the system

ẋ = f (t, x) if
x(t0 ) = x∗ → x(t) ≡ x∗ , ∀t ≥ t0
is satisfied.

✔ The equilibrium points of autonomous systems in the form ẋ = f (x) are the real
solutions of
f (x) = 0.

✔ Similarly, if a vector x∗ at time t0 and for all t ≥ t0 satisfies

f (t, x∗ ) = 0

it is a equilibrium point of the system ẋ = f (t, x).

✔ For autonomous systems if a point x∗ is an equilibrium point at some time instant
t0 , it is an equilibrium point for all time instants.
 Nonlinear Systems 12

✔ Equilibrium points can be isolated or a continuum of equilibrium points.

✔ The equilibrium points of linear systems in the form ẋ = Ax
➙ Could be single/unique (isolated) point x∗ = 0 (if matrix A is nonsingular)
➙ Could be a continuum of equilibrium points passing through point x∗ = 0 (if
matrix A is singular every point in the null space of A is an equilibrium point)
(equilibirum sub-space)

➙ It is impossible to have multiple isolated equilibrium points since if xa and xb

are equilibirium points, from the linearity property all points on the line
αxa + (1 − α)xb are also equilibrium points.
 Nonlinear Systems 13

✔ Nonlinear systems may have multiple isolated equilibrium points.

➙ For example, the system

 ẋ
1 = x2 ,
 ẋ2 = −a sin(x1 ) − bx2

has infinitely many isolated equilibrium points at

(x1 = nπ, x2 = 0), n = 0, ±1, ±2, ....
✔ Linearization of nonlinear systems in general is performed about equilibrium points
or operating points.
 Nonlinear Systems 14

Linearization

✔ In analysis of nonlinear systems it is very common to linearize the system about

an operating or an equilibrium point and analyze the resultant linear equivalent
system

➙ Allows to utilize methodologies for analysis of linear systems

➙ Allows to determine the local behavior and local qualitative properties of the
system (to some extend)

✔ Limitations of linearization
➙ Using linearization one can determine only local properties/behavior of the
system. Global beavior/properties cannot be determined.

➙ The results obtained through linearization may not always hold for the nonlinear
system.

➙ There are purely nonlinear properties/behavior which cannot be seen in linear

systems. Such properties/behavior cannot be observed in linearized systems.
 Nonlinear Systems 15

Purely Nonlinear Phenomena

✔ Finite escape time: The state/output of system may diverge to infinity in finite time
✔ Multiple isolated equilibrium points: In linear system the unique isolated
equilibrium point can be x∗ = 0. In nonlinear systems there may be multiple
isolated equilibrium points.

✔ Limit cycles: Isolated periodic solutions/orbits

✔ Subharmonic, harmonic, or almost-periodic oscillations: In linear systems, the
frequency of the state and output is equal to the frequency of the input. In
nonlinear systems, frequencies which are multiples of the input frequency are
possible.

✔ Chaos: Even though there are no probabilistic components, the system can lead
to responses which appear as probablistic.

✔ Multiple modes of behavior: The same system may possess multiple different
qualitative properties simultaneously. The system may have multiple equilibria,
limit cycles and harmonic oscillations in some regions.