Nonlinear Systems and Control

Lecture 2 (Meeting 5)

Eugenio Schuster

[email protected]
Mechanical Engineering and Mechanics
Lehigh University

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 1 / 17
Nonlinear Models
In this course we will deal with nonlinear dynamical systems that are model by a
set of coupled first-order ordinary differential equations (ODE),

ẋ1 = f1 (t, x1 , . . . , xn , u1 , . . . , up )
ẋ2 = f2 (t, x1 , . . . , xn , u1 , . . . , up )
.. . (1)
. = ..
ẋn = fn (t, x1 , . . . , xn , u1 , . . . , up )

where x1 , . . . , xn denote the n states, u1 , . . . , up denote the p inputs, t denotes

time and ẋi denotes the time derivative of the state xi .

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 2 / 17
Nonlinear Models
After defining
     
x1 u1 f1 (t, x, u)
 x2   u2   f2 (t, x, u) 
x= , u= , f (t, x, u) = 
     
.. .. .. 
 .   .   . 
xn up fn (t, x, u)

we can rewrite the state equation (1) as

ẋ = f (t, x, u) (2)

which may be associated with the output equation

y = h(t, x, u) (3)

where y denotes the q-dimensional output.

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 3 / 17
Nonlinear Models

Nonlinear Control: Design control law

u = γ(t, x)

for
ẋ = f (t, x, u)
Nonlinear Analysis: We study the dynamics of the unforced system

ẋ = f (t, x)

where u has been either forced to zero or replaced by the control law γ(t, x).

ẋ = f (t, x) nonautonomous or time-varying

ẋ = f (x) autonomous or time-invariant

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 4 / 17
Nonlinear Models
A point x = x∗ in the state space is said to be an equilibrium point if it has the
property that whenever the state of the system starts at x∗ , it will remain at x∗
for all future time. For the autonomous system

ẋ = f (x) (4)

the equilibrium points are the real roots of the equation

0 = f (x) (5)

Equilibrium points can be isolated or there can be a continuum of points.

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 5 / 17
Nonlinear Phenomena
How nonlinear systems are different from our well-known linear systems?
Multiple isolated equilibria
Finite escape time
Limit cycles
Chaos
Subharmonic, harmonic or almost periodic oscillations
Multiple modes of behavior

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 6 / 17
Nonlinear Phenomena
Multiple isolated equilibria: For linear time-invariant (LTI) systems

ẋ = Ax

the equilibria are given by the null space of A, N (A). A linear system can have
only one isolated equilibrium point (A is full rank). A nonlinear system can have
more than one isolated equilibrium point.
Examples:
ẋ = −x + x3
The points x = 0 (stable) and x = ±1 (unstable) are isolated equilibrium points.

ẋ = x − x3

The points x = 0 (unstable) and x = ±1 (stable) are isolated equilibrium points.

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 7 / 17
Nonlinear Phenomena
The solution of
ẋ = x − x3
is given by
x0
x(t) =
x20 (1 − e−2t ) + e−2t
2

1.5

0.5
x(t)

−0.5

−1

−1.5

−2
0 2 4 6 8 10
t

Figure: Multiple equilibria

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 8 / 17
Nonlinear Phenomena
Finite escape time: For linear time-invariant (LTI) unstable systems

ẋ = Ax

the state goes to infinity as time approaches infinity. For nonlinear systems, the
state can go to infinity in finite time.
Examples:
ẋ = x
with solution
x(t) = x0 et

ẋ = x3
with solution
x0
x(t) = p
1 − 2x20 t

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 9 / 17
Nonlinear Phenomena
x0
Note that x(t) = x0 et goes to infinity when t → ∞, while x(t) = √ goes
1−2x20 t
1
to infinity when t → 2x20
.

15 20

15
10

10

5
5
x(t)

x(t)
0 0

−5
−5

−10

−10
−15

−15 −20
0 0.5 1 1.5 2 0 0.5 1 1.5 2
t t

t
Figure: Linear system (left) with solution x(t) = x0 e vs. Nonlinear system (right) with
solution x(t) = √ x0 2 .
1−2x0 t

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 10 / 17
Nonlinear Phenomena
Limit Cycles: For linear time-invariant (LTI) system to oscillate, it must have a
pair of eigenvalues on the imaginary axis, which is a nonrobust condition that is
almost impossible to maintain in the presence of perturbations. Even if we do, the
amplitude of the oscillation will depend on the initial state.
In real life, stable oscillators must be produced by nonlinear systems.
There are nonlinear systems that can go into an oscillation of a fixed
amplitude and frequency, irrespective of the initial state.
This type of oscillation is known as a limit cycle (isolated periodic orbit).

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 11 / 17
Nonlinear Phenomena
Examples:
Linear: ẍ + x = 0 Nonlinear: ẍ + (x2 − 1)ẋ + x = 0

2 5

4
1.5
3
1
2
0.5
1
x2

x2
0 0

−1
−0.5
−2
−1
−3
−1.5
−4

−2 −5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −8 −6 −4 −2 0 2 4 6 8
x1 x1

Figure: Linear system (left) vs. Nonlinear system (right)

For the limit cycle (right figure), the damping is positive for large x and negative
for small x. This is a version of the well-known Van der Pol equation.

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 12 / 17
Nonlinear Phenomena
Lemma 2.1: (Poincaré-Bendixson Criterion) Consider the system

ẋ = f (x) (6)

and let M be a closed bounded subset of the plane such that

M contains no equilibrium points, or contains only one equilibrium point such
that the Jacobian matrix [∂f /∂x] at this point has eigenvalues with positive
real parts. (Hence, the equilibrium point is unstable focus or unstable node.)
Every trajectory, starting in M stays in M for all future time.
Then, M contains a periodic orbit of (6). The lemma guarantees existence but
not uniqueness.

Figure: Redefinition of M to exclude unstable node or focus

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 13 / 17
Nonlinear Phenomena
As a tool for investigating whether trajectories are trapped inside a set M , let us
consider a simple closed curve defined by the equation V (x) = c, where V (x) is
continuously differentiable. The vector field f (x) at a point x on the curve points
inward if the inner product of f (x) and the gradient vector ∇V (x) is negative, i.e.

∂V ∂V
f (x) · ∇V (x) = (x)f1 (x) + (x)f2 (x) < 0. (7)
∂x1 ∂x2
Let us consider the following cases:
For a set of the form M = {x : V (x) ≤ c} for some c > 0, trajectories are
trapped inside M if f (x) · ∇V (x) ≤ 0 on the boundary V (x) = c.
For a set of the form M = {x : W (x) ≥ c1 and V (x) ≤ c2 } for some c1 > 0
and c2 > 0 (annular region), trajectories are trapped inside M if
f (x) · ∇V (x) ≤ 0 on the boundary V (x) = c2 and f (x) · ∇W (x) ≥ 0 on the
boundary W (x) = c1 .

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 14 / 17
Nonlinear Phenomena

If the derivative dV ∂V
dt = ∂x f (x) along a phase trajectory is everywhere negative,
then the trajectory tends to the origin (inward).

V̇ ≡ dV
dt will be negative as long as the angle ϕ between grad V ≡
∂V
∂x and
◦
ẋ ≡ dx
dt = f (x) is higher than 90 .

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 15 / 17
Nonlinear Phenomena
Example: Harmonic oscillator – ẍ + x = 0

ẋ1 = x2
ẋ2 = −x1

1.5

0.5
x2

−0.5

−1

−1.5

−2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
x1

Figure: Limit cycle

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 16 / 17
Nonlinear Phenomena
Example:

ẋ1 = x1 + x2 − x1 (x21 + x22 )

ẋ2 = −2x1 + x2 − x2 (x21 + x22 )

1.5

0.5
x2

−0.5

−1

−1.5

−2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
x1

Figure: Limit cycle

Prof. Eugenio Schuster ME 450 - Nonlinear Systems and Control Spring 2024 17 / 17