EE250: Lecture Note

Nonlinear Systems and Linearization

1 Nonlinear Systems
Model:

ẋ = f (t, x, u)
y = h(t, x, u)

Properties:

• Don’t follow the principle of superposition, i.e,

h(t, x, a1 u1 + a2 u2 ) 6= a1 h(t, x1 , u1 ) + a2 h(t, x2 , u2 )

• Multiple equilibrium points

• Limit cycles : oscillations of constant amplitude and frequency
• Subharmonic, harmonic oscillations for constant frequency inputs
• Chaos: randomness, complicated steady state behaviours
• Multiple modes of behaviour

Terminology:

Autonomous Systems: the nonlinear function does not explicitly depend on time t.

ẋ = f (x, u)
y = h(x, u)

Affine System:
ẋ = f (x) + g(x)u
Unforced System: input u(t) = 0,
ẋ = f (x)

Example:

1
 Pendulum:
b

ml2 θ̈ = −mglsin(θ) − k θ̇

The state equations are θ

ẋ1 = x2 l
g k b
ẋ2 = − sin(x1 ) − 2 x2
l ml
mg
1.0.1 Exercise
Identify the category to which the following differential equations belong to? Why?
1. ẋ = tx
2. ẋ = −x + x2
3. ẋ = ux where u is an external input.
4. ẋ = −2x + t2
5. ẋ = −x

1.1 Linearization
• Concept of Equilibrium Point: Consider a system
ẋ = f (x, u)
where functions fi (.) are continuously differentiable. The equilibrium point (xe , ue ) for
this system is defined as
f (xe , ue ) = 0

• What is linearization?
Linearization is the process of replacing the nonlinear system model by its linear coun-
terpart in a small region about its equilibrium point.
• Why do we need it?
We have well stablished tools to analyze and stabilize linear systems.

1.1.1 The method:

Let us write the general form of nonlinear system ẋ = f (x, u) as:
dx1
= f1 (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )
dt
dx2
= f2 (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )
dt
.. .
. = ..
dxn
= fn (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )
dt

2
Let ue = [u1e u2e . . . ume ]T be a constant input that forces the system ẋ = f (x, u) to
settle into a constant equilibrium state xe = [x1e x2e . . . xne ]T such that f (xe , ue ) = 0
holds true.
We now perturb the equilibrium state by allowing: x = xe + ∆x and u = ue + ∆u.
Taylor’s expansion yields
dx ∂f ∂f
= f (xe + ∆x, ue + ∆u) = f (xe , ue ) + (xe , ue )∆x + (xe , ue )∆u + ... (1)
dt ∂x ∂u
where
 ∂f1 ∂f1
  ∂f1 ∂f1

∂x1
··· ∂xn

∂u1
··· ∂um

∂f ∂f
(xe , ue ) =  ... ..  (xe , ue ) =  ... .. 
 
(2)
 
∂x .  ∂u . 
∂fn ∂fn  ∂fn ∂fn 
∂x1
··· ∂xn xe ,ue ∂u1
··· ∂um xe ,ue

are the Jacobian matrices of F with respect to x and u, evaluated at the equilibrium point,
[xTe uTe ]T .
Note that
dx dxe d(∆x) d(∆x)
= + = (3)
dt dt dt dt
because xe is constant. Furthermore, f (xe , ue ) = 0. Let

∂f ∂f
A= (xe , ue ) and B = (xe , ue ) (4)
∂x ∂u
Neglecting higher order terms, we arrive at the linear approximation

d(∆x)
= A∆x + B∆u (5)
dt
Similarly, if the outputs of the nonlinear system model are of the form

y1 = h1 (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )
y2 = h2 (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )
.. .
. = ..
yp = hp (x1 , x2 , . . . , xn , u1 , u2 , . . . , um )

or in vector notation
y = h(x, u) (6)
then Taylor’s series expansion can again be used to yield the linear approximation of the
above output equations. Indeed, if we let

y = y e + ∆y (7)

then we obtain
∆y = C∆x + D∆u (8)

3
Example 1. Consider a first order system:

ẋ = −x + x2 where x(0) = x0

Linearize it about origin

ẋ = −x
Its solution is : x(t) = x0 e−t . Whatever may be the initial state x0 , the state will settle
at x(t) = 0 as t− > ∞, which is the only equilibrium point that this linearized system
has.
However, the solution of actual nonlinear system is

x0 e−t
x(t) =
1 − x0 + x0 e−t
For various initial conditions, the system has two equilibrium points: x = 0 and x = 1
as can be seen in the figure 1.1.1.

Figure 1: Multiple equilibrium points

Example 2. Consider a nonlinear system

−1
ẋ1 = (9a)
x22 (t)

ẋ2 = u(t)x1 (t) (9b)

4
 Linearize the system (9) about the nominal trajectory [x01 (t), x02 (t)], which is the
solution to the equations with initial condition x1 (0) = x2 (0) = 1 and input u(t) = 0.
Integrating both sides of Eqn. (9b) with respect to t, we have,

x2 (t) = 1

Now, Eqn. (9a) gives

x1 (t) = −t + 1
Therefore, the nominal trajectory about which Eqns. (9a) and (9b) are to be linearized
is described by

x01 (t) = 1 − t
x02 (t) = 1 (10)

Now, evaluating the coefficients of Eqn. (4), we get

∂f1 (t) ∂f1 (t) 2 ∂f2 (t) ∂f2 (t)

=0 = 3 = u(t) = x1 (t)
∂x1 (t) ∂x2 (t) x2 (t) ∂x1 (t) ∂x2 (t)

Hence, the linearized equations are given by

      
△ẋ1 (t) 0 2 △x1 (t) 0
= + △u(t) (11)
△ẋ2 (t) 0 0 △x2 (t) 1−t

which is a set of linear state equations with time-varying coefficients.

1.1.2 Exercise
1. Consider a nonlinear system

ẋ1 = −x21 + x1 x2
ẋ2 = −2x22 + x2 − x1 x2 + 2
 
1
(a) Show that xe = is an equilibrium state.
1
(b) Is it the only equilibrium state? What are the others, if any?

2. You are given a nonlinear input-output system which satisfies the nonlinear differential
equation:
d2 y dy du
2
= 2y − (y 2 + 1)( + 1) + u + 2
dt dt dt
(a) Obtain a nonlinear state-space representation.
(b) Linearize this system around its equilibrium point when u(.) = 0.

[Hint: Select states as x1 = y; x2 = ẏ − 2u.]