Nonlinear Control Systems

Dr. Muhammad Abid

1
 Why nonlinear control ?
Linearized models become less accurate for
Improved control wider operating range
𝑥=
˙ 𝑓 ( 𝑥)
Examples

• In robotic systems, many

𝑑𝑓
forces vary as square of ≈ 𝑓 ( 𝑥0 ) + | ( 𝑥 − 𝑥0)
𝑑𝑥 𝑥= 𝑥 0

speed. Linear models

become less accurate as 𝑑𝑓
𝑓 (𝑥 )− 𝑓 ( 𝑥𝑜) ≈ | ( 𝑥 − 𝑥0)
speed increases. 𝑑𝑥 𝑥=𝑥 0

• In liquid level systems, Define , and

flow rates depend upon
square root of liquid head. 𝑑𝑓
𝛿 𝑥˙ ≈ | 𝛿𝑥
𝑑𝑥 𝑥= 𝑥 0

2
Why nonlinear control ?
Many nonlinearities cannot be linearly
Improved control approximated
Hard nonlinearities
• Saturation nonlinearity
• Dead-zone nonlinearity
• Backlash nonlinearity
• Hysteresis

3
Why nonlinear control ?
Nonlinearities are intentionally added in
Improved control controller to tolerate uncertainties
Hard nonlinearities
• ON-OFF control
Dealing with uncertainties • Adaptive control

4
Why nonlinear control ?
Improved control
Hard nonlinearities
Dealing with uncertainties
Less expensive Sensors and actuators with linear
behavior are more expensive

Modern computers have made it easier to implement

nonlinear controllers

5
Types of nonlinearities
 Inherent Nonlinearites
 Centripetal forces
Inherent or Intentional  Columb friction

 Intentional nonlinearites
 ON-OFF control
 Adaptive control

6
Types of nonlinearities
 Continuous nonlinearities
 Square, square-root, multiplication of variables, etc.
Inherent or Intentional
Example:
Continuous or Discontinuous
𝑑h
𝐴 =𝑄𝑖 − 𝑐 √ h
𝑑𝑡

 Dis-continuous nonlinearities (hard nonlinearities)

 Dead-zone
 Saturation
 ON-OFF control
 Cannot be locally approximated
 Have severe effect on stability

7
Linear systems vs Nonlinear systems
Equilibrium point: if state of a system starts
Linear Systems at , it will remain at for future time
 Modelled by linear differential equations ⇒
0= 𝐴 𝑥 𝑒𝑞 ❑ 𝑥 𝑒𝑞 =0
𝑥=
˙ 𝐴𝑥

𝑥(𝑡 )=𝑥 ( 0 ) 𝑒 𝐴𝑡
 Superposition holds
 Unique equilibrium point
 Stability is independent of initial conditions
 Stability is independent of input, asymptotic 𝑥=
˙ 𝐴𝑥+ 𝐵𝑢
stability of implies BIBO stability of 𝑡
𝑥 ( 𝑡 ) =𝑥 ( 0 ) 𝑒 +∫ 𝑒
𝐴𝑡 𝐴 ( 𝑡 −𝜏 )
 For sinusoidal inputs, the output is 𝐵𝑢 ( 𝜏 ) 𝑑 𝜏
0
sinusoidal
8
Nonlinear Systems
Example: Under-water vehicle
1

Superposition may not be

𝑣˙ +|𝑣|𝑣 =𝑢
applicable 0.5

0
0 5 10 15 20
• Settles faster in response to positive step
• Scaling input does not result into the same scaling in
output
4
𝑢=1
⇒ 0 +|𝑣 𝑠|𝑣 𝑠 =1 3

⇒ 𝑣 𝑠=1 2

𝑢=10 1

0
0 5 10 15 20
9 ⇒𝑣 𝑠= √ 10=3.16
 Nonlinear Systems
Example: Pendulum
Superposition may not be
applicable
𝑥˙ 1=𝑥 2
Multiple isolated • Set and
equilibrium points
0=𝑥 2
Linear systems may also have multiple
• Which gives and
equilibrium points but these are not isolated
𝑥2
• Equilibrium points
𝑥˙ 1=𝑥 2 𝑥𝑒𝑞 =( 𝑥 1 ,0) 𝑥1 ⋯ , ( − 𝜋 , 0 ) , ( 0,0 ) , ( 𝜋 ,0 ) , ( 2 𝜋 , 0 ) , ⋯

10
Nonlinear Systems
Example:
Superposition may not be 𝑥=−
˙ 𝑥+ 𝑥 2
applicable
• Two equilibrium points and
Multiple isolated • Response for different initial conditions
equilibrium points 2
Stability may depend upon 1.5
initial conditions 1

0.5

-0.5

-1
0 2 4 6 8 10
11
Nonlinear Systems
Superposition may not be
applicable
Multiple isolated
equilibrium points
Stability may depend upon
initial conditions Example:
Stability may depend upon 𝑥=𝑥𝑢
˙
input • For , , and converges to zero
• For , , and tends to infinity

12
Nonlinear Systems
Superposition may not be
applicable
Multiple isolated • Unstable linear systems: infinite time for
equilibrium points the response to reach to infinity
Stability may depend upon 𝑥=𝑥
˙ ⇒ 𝑥= 𝑥 ( 0 ) 𝑒𝑡
initial conditions
Stability may depend upon • Nonlinear systems: States may go to
infinity in a finite time
input
1
Finite escape time 𝑥=−
˙ 𝑥2 ⇒ 𝑥=
𝑡 −1

13
Nonlinear Systems
Superposition may not be
applicable
Multiple isolated
equilibrium points • Linear systems: A periodic input produces
an output of the same frequency.
Stability may depend upon
initial conditions
Stability may depend upon • Nonlinear systems: Can produce outputs
with frequencies which are multiples or
input sub-multiples of input frequency
Finite escape time
Subharmonic, superharmonics
14
 Linear Systems:
Limit cycle 5
• Eigenvalues on imaginary axis will
result into oscillatory response
Limit cycles are
0
˙
[ 01
𝑥=
−4
0
𝑥
]
oscillations of fixed
• Non-robust oscillations, difficult to keep
amplitude and • Amplitude of oscillations depends upon
period -5 ICs
0 5 10 15 20

Nonlinear Systems may exhibit limit cycles 6

Van der Pol equation

4
¨ +2 𝑐 ( 𝑥 − 1 ) 𝑥+𝑘𝑥
2
𝑚𝑥 ˙ =0 𝑘
2
• Mass-spring-damper with position
dependent damping 0
• RLC circuit with nonlinear resistor
• Sustained oscillations, -2
2
• Do not depend upon ICs 𝐷 ( 𝑥 ) =2 𝑐( 𝑥 − 1)
• Desirable/undesirable -4
15 0 20 40 60 80 100
Bifurcation
Duffing equation
Change in 𝑥+𝛼
¨ 𝑥+ 𝑥 3= 0
parameters  Change in number of
0+𝛼 𝑥 +𝑥 3 =0
equilibrium points 𝑥 ( 𝛼 + 𝑥 )= 0
2
One
 Different behavior of equilibrium
𝛼> 0 , ⇒ 𝑥=0 , 𝑥=± 𝑗 √ 𝛼 point
each equilibrium
Three
point 𝛼< 0 , ⇒ 𝑥=0 , 𝑥=± √ −𝛼 equilibrium
points

Bifurcation: Quantitative change in 𝑥𝑒𝑞

parameter resulting into qualitative Pitchfork
change of system properties Bifurcation
𝛼

16
Chaos
𝑥+0.1
¨ ˙ 𝑥 5=6 sin 𝑡
𝑥+

Behavior of nonlinear 4

systems may be
𝑥 ( 0 )=2
2

extremely sensitive to 0

small changes in initial 𝑥 ( 0 )=2.01 -2

conditions / input /
parameters -4
0 5 10 15 20 25 30

Different from 3

stochasticity / 6 sin 𝑡
2

1
randomness 5.99 sin 𝑡 0

-1

-2

-3
17 0 5 10 15 20 25 30
Course contents
Pre-requisites
Nonlinear systems Analysis
Phase plane analysis 1. A course on linear systems
Lyapunov stability 2. Differential equations
Describing function analysis
Nonlinear systems design
Feedback linearization based design
Sliding mode control
Lyapunov redesign References
Backstepping
1. H. Khalil, “Nonlinear Systems”, 3rd edition
2. J.J. Slotine, “Applied Nonlinear Control”
3. A. Isidori, “Nonlinear Control Systems”
18