Nonlinear Control

Chapter 7: Nonlinear Controller

Design
In this chapter we review a few nonlinear control Lyapunov-
based design methods. Backstepping method is introduced
first, while Sliding mode control, as one of the most celebrated
robust control schemes are reviewed next. Finally Lyapunov
redesign method is described, to robustifying a nominal
nonlinear controller.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 Welcome
To Your Prospect Skills
On Nonlinear System Analysis and
Nonlinear Controller Design . . .

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 Get more global exposure
Self confidence is the first key

About ARAS
ARAS Research group originated in 1997 and is proud of its 22+ years of brilliant background, and its contributions to
the advancement of academic education and research in the field of Dynamical System Analysis and Control in the
robotics application.ARAS are well represented by the industrial engineers, researchers, and scientific figures graduated
from this group, and numerous industrial and R&D projects being conducted in this group. The main asset of our
research group is its human resources devoted all their time and effort to the advancement of science and technology.
One of our main objectives is to use these potentials to extend our educational and industrial collaborations at both
national and international levels. In order to accomplish that, our mission is to enhance the breadth and enrich the
quality of our education and research in a dynamic environment.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 4
Contents

Backstepping
1 Design idea, integrator and general backstepping, recursive backstepping, examples.

Sliding Mode Controller

2
Motivating example, chattering, continuous SMS, integral SMC. Single-input SMC; tracking
objective, sliding manifold, variable structure control, continuous SMC, robust IOFL. Multi-Input
SMC; regular form, sliding condition, chattering reduction, UUB stability, unmatched uncertainty.

Lyapunov Redesign
3 Robust stabilization, nominal Lyapunov function, 2-norm and ∞-norm perturbation, boundary
layer, nonlinear damping, examples.

In this chapter we review a few nonlinear control Lyapunov-based design methods. First backstepping method is introduced to iteratively
generate a complex Lyapunov function from low order subsystems. Sliding mode control, as one of the most celebrated robust control schemes
are reviewed for single-input and multi-input systems. Finally, Lyapunov redesign is described, on how a robust stabilizing controller can be
generated by robustifying a nominal nonlinear controller.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 5
Nonlinear Controller Design
• Backstepping
 Design Idea: Special Case
• Integrator Backstepping
𝜂ሶ = 𝑓 𝜂 + 𝑔 𝜂 𝜉 (7.1)
𝜉ሶ = 𝑢 (7.2)
• Design a state feedback to stabilize the origin (𝜂 = 0, 𝜉 = 0)

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 6
Backstepping
• Design Idea:
 Stabilize 7.1 by 𝜉 = 𝜙(𝜂), with 𝜙 0 = 0.
• The system is: 𝜂ሶ = 𝑓 𝜂 + 𝑔 𝜂 𝜙(𝜂)
• 𝜂 = 0 is the origin
– Seek for any asymptotically stabilizing controller 
• For a Lyapunov Candidate 𝑉(𝜂), 𝑉(0) = 0.
• The derivative is negative definite.

 Extend to overall system 7.1 and 7.2 by

• Add and subtract 𝑔 𝜂 𝜙(𝜂) on the RHS of 7.1

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 7
Backstepping
• Design Idea (Cont.)
 Block diagram

• Use change of variable: 𝑧 = 𝜉 − 𝜙 𝜂

– state equation: 𝜂ሶ = 𝑓 𝜂 + 𝑔 𝜂 𝜙 𝜂 =𝑔 𝜂 𝑧
𝑧ሶ = 𝑢 − 𝜙ሶ

• Integrator backstepping:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 8
Backstepping
• Design Idea (Cont.)
• Since 𝑓, 𝑔, and 𝜑 are known:
• Take 𝑣 = 𝑢 − 𝜙ሶ to reach to cascade connection:
𝜂ሶ = 𝑓 𝜂 + 𝑔 𝜂 𝜙 𝜂 =𝑔 𝜂 𝑧
𝑧ሶ = 𝑣
• This is similar to 7.1 and 7.2 !
𝜂ሶ = 𝑓 𝜂 + 𝑔 𝜂 𝜉
𝜉ሶ = 𝑢
– But the origin is asymptotically stable with zero input.
• Use this feature to design for 𝑣:
– Lyapunov Candidate for overall system:
– Its derivative:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 9
Backstepping
• Design Idea (Cont.)
• Choose:

– Yields to:

– The origin (𝜂 = 0, 𝜉 = 0) is asymptotically stable.

• Substitute:
– The stabilizing controller:

– If all assumptions holds globally and V(η) is radially unbounded, the

origin is globally asymptotically stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 10
Backstepping
• Integrator Backstepping
 Example 1:
• Consider the system

– Set the variables: 𝜂 = 𝑥1 , 𝜉 = 𝑥2 .

– Start with scalar equation:

– Choose a stabilizing feedback control

– This yields to:

1
– Lyapunov Candidate: 𝑉 𝑥1 = 𝑥12
2
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 11
Backstepping
• Integrator Backstepping
 Example 1: (Cont.)
– Lyapunov Derivative:

– The origin is globally exponentially stable.

– To backstep:

– The system changes to:

– Take the composite Lyapunov function as:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 12
Backstepping
• Integrator Backstepping
 Example 1: (Cont.)
– The Lyapunov function derivative:

– By taking the control input as:

– This yields to:

– Hence, the origin is globally asymptotically stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 13
Backstepping
• Integrator Backstepping
 Example 2: Recursive Backstepping
• System Dynamics:

– 2nd order system

– Can be stabilized by:

– And
– 3rd order system viewed as:

– Change of variable: 𝑧3 = 𝑥3 − 𝜑(𝑥1 , 𝑥2 )

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 14
Backstepping
• Integrator Backstepping
 Example 2: Recursive Backstepping (cont.)
– Overall stabilizing control law:

– Overall Lyapunov function

1 2
𝑉𝑐 𝑥 = 𝑉 𝑥1 , 𝑥2 + 𝑥3 − 𝜙 𝑥1 , 𝑥2
2
– Its derivative is:

– Hence, the origin is globally asymptotically stable

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 15
Backstepping
• General Backstepping
 System Dynamics:

– In which 𝑔𝑎 𝜂, 𝜉 ≠ 0
• Input transformation:

– Reduces the last equation to 𝜉ሶ = 𝑢𝑎

• Controller Input:

• For some 𝑘 > 0:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 16
Backstepping
• Recursive Backstepping
 Strict Feedback System:

– Assume that:
• Recursive procedure
– Start with: where,
– Find stabilizing controller
– With Lyapunov derivative:
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 17
Backstepping
• Recursive Backstepping
• Recursive procedure (cont.)
– Move a step forward

– Set:
– Generate control law and Lyapunov function:

– 3rd Step:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 18
Backstepping
• Recursive Backstepping
• Recursive procedure (cont.)
– Set:

– Generate control law

– For some 𝑘 > 0

– Lyapunov function:

– Repeat k times to find the stabilizing controller:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 19
Backstepping
• Recursive Backstepping
• Example 1:
– Consider Strict Feedback
Input-output normal form:

– In which 𝑥 defines the internal dynamics and is assumed to be in

strict feedback form.
– If the system is minimum phase the origin of the system 𝑥0 = 𝑓0 (𝑥) is
globally asymptotically stable.
– We can find a Lyapunov function 𝑉0 that satisfies:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 20
Backstepping
• Recursive Backstepping
• Example 1: (Cont.)
– If the system is minimum phase:
• Choose 𝜑0 (𝑥) = 0, and recursively find 𝑢.

– If the system is non-minimum phase:

• Choose 𝜑0 𝑥 , 𝑉0 (𝑥) such that the non-minimum phase Internal dynamics
is stabilized.
• Recursively find the stabilizing controller.

– Conclusion:
•  always a stabilizing controller for minimum phase systems
•  a stabilizing controller for non-minimum phase systems, if we could find
a stabilizing 𝜑0 (𝑥) for the internal dynamics.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 21
Backstepping
• Recursive Backstepping
• Example 2:
– Consider the 2nd order system:

– Start with the first equation, and choose

– Lyapunov derivative:

– Design Backstepping controller and Lyapunov function:

and
– Lyapunov derivative is

– The eq. point is globally asymptotically stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 22
Backstepping
• Example 3:
– Consider the 2nd order system:

– The first equation, cannot be stabilized by

– Choose
– Lyapunov derivative:

– Design Backstepping controller

– and Lyapunov function:

– Lyapunov derivative is
– The eq. point is globally asymptotically stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 23
Backstepping
• Example 4:
– Consider the system:
– where, 𝛼, 𝛽, 𝜇 are known constants
– The system is in standard form, therefore,
–
– In which: and

– Use the standard a stabilizing controller

– Start with the stabilizing controller for the first equation:

– The stabilizing controller can be find by:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 24
Backstepping
• Example 5:
– Consider the system:

– Start with the first equation:

– Choose a stabilizing controller

– Now check the first two equations:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 25
Backstepping
• Example 5: (Cont.)
– The Lyapunov equation is:

– At the first stage:

– And the last step:

– The final Lyapunov function can be found by

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 26
Backstepping
• Example 6:
– Consider the system:

– At the first stage:

– One step backing:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 27
Backstepping
• Example 6: (Cont.)
– Consider the system:

– At the first stage:

– One step backing:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 28
Backstepping
• Example 6: (Cont.)
– Another step back:

– In which,

– And the final Lyapunov function is:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 29
Contents

Backstepping
1 Design idea, integrator and general backstepping, recursive backstepping, examples.

Sliding Mode Controller

2
Motivating example, chattering, continuous SMS, integral SMC. Single-input SMC; tracking
objective, sliding manifold, variable structure control, continuous SMC, robust IOFL. Multi-Input
SMC; regular form, sliding condition, chattering reduction, UUB stability, unmatched uncertainty.

Lyapunov Redesign
3 Robust stabilization, nominal Lyapunov function, 2-norm and ∞-norm perturbation, boundary
layer, nonlinear damping, examples.

In this chapter we review a few nonlinear control Lyapunov-based design methods. First backstepping method is introduced to iteratively
generate a complex Lyapunov function from low order subsystems. Sliding mode control, as one of the most celebrated robust control schemes
are reviewed for single-input and multi-input systems. Finally, Lyapunov redesign is described, on how a robust stabilizing controller can be
generated by robustifying a nominal nonlinear controller.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 30
Sliding Mode Control
• Sliding Mode Control (SMC)
 Robust Controller (Lyapunov Based)

• System with modeling uncertainty

– Structured or parametric uncertainty
– Unstructured uncertainty
– Additive or multiplicative
– Matched or un matched

• Variable Structure Control (VSC)

– Two phase controller
– Switching structure

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 31
Sliding Mode Control
• Motivating Example
 Consider SI system:
• Where ℎ and 𝑔 are uncertain functions:
– and 𝑔 𝑥 ≥ 𝑔0 > 0, ∀𝑥.
• Suppose with the input 𝑢 the system is enforced to the following
stable manifold: 𝑠 = 𝑎1 𝑥 + 𝑥2 = 0
– By choosing 𝑎1 > 0 the trajectories 𝑥 𝑡 → 0 as 𝑡 → ∞.
– The motion on 𝑠 = 0 is independent of ℎ, 𝑔.
• How can we bring the trajectories to 𝑠 = 0, and maintain it there?
1
– Use Lyapunov-based control with 𝑉 𝑠 = 𝑠 2
2
– Geometrically: (𝑠 2 ) is the squared distance to the surface 𝑠.
– Check the convergence: 𝑉ሶ 𝑠 = 𝑠 𝑠ሶ
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 32
Sliding Mode Control
• Motivating Example
• How can we bring the trajectories to 𝑠 = 0, and maintain it there?
𝑠ሶ = 𝑎1 𝑥ሶ 1 + 𝑥ሶ 2 = 𝑎1 𝑥2 + ℎ 𝑥 + 𝑔 𝑥 𝑢
– But the functions ℎ, and 𝑔 are uncertain, and not known exactly!
– Suppose ℎ, and 𝑔 are bounded by:

– Where 𝜚 is some known function. Hence,

– To make 𝑉ሶ negative definite use a switching controller:

– where,

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 33
Sliding Mode Control
• Motivating Example
• How can we bring the trajectories to 𝑠 = 0, and maintain it there?
– This controller is very effective since 𝑉ሶ becomes negative definite:
.

– In which, 𝛽 is selected such that 𝛽 𝑥 ≥ 𝜚 + 𝛽0 . Let 𝜂 = g 0 𝛽0 > 0:

• Furthermore it results into finite time reaching phase:

𝑉ሶ ≤ −𝜂 𝑠 Or 𝑉ሶ ≤ −𝜂 2𝑉 ⟹ 𝑠 𝑡 ≤ 𝑠 0 −𝜂𝑡
→ t 𝑟𝑒𝑎𝑐ℎ ≤ 𝑠 0 /𝜂

– Thus, 𝑠 𝑡 reaches to 𝑠 = 0 manifold in finite time.

– On the surface 𝑠 = 0, the trajectory moves toward the origin.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 34
Sliding Mode Control
• Motivating Example
• In summary
– A switching and Lyapunov-based controller:
1
𝑢 = −𝛽 sgn(𝑠), and 𝑉 𝑠 = 𝑠2
2
– Then 𝑉ሶ ≤ −𝜂 𝑠 for 𝜂 > 0.
• This is called sliding condition.
• The system response has two phases:
– Reaching phase: to reach to the sliding manifold (𝑠 = 0) in finite time.
– Sliding phase: To slide along the sliding manifold to the origin.
– The dynamics of the system in sliding phase is: 𝑥ሶ1 = −𝑎1 𝑥2 .
– This is illustrated in the figure.
– Then such 𝑢 is called a sliding mode control (SMC).
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 35
Sliding Mode Control
• Motivating Example
• The SMC simplifies if

Then
This is a simple relay.

 Chattering
• Discontinuous Control across s(t)
• Imperfect Control
– Discrete time OR Quantization error
– Chattering: Oscillation across s(t)

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 36
Sliding Mode Control
 Chattering
• Example: Consider SMC for pendulum as:
𝜋
– To stabilize the pendulum at 𝛿1 = , choose 𝑥1 = 𝜃 − 𝛿1 , 𝑥2 = 𝜃ሶ
2

– Assume parametric uncertainties:

0.05 ≤ 𝑚 ≤ 0.2, 0.9 ≤ ℓ ≤ 1.1, 0 ≤ 𝑘0 ≤ 0.05
– Then the SMC gain may be selected as 𝑘 = 0.4, since:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 37
Sliding Mode Control
 Chattering
• Example: (Cont.)
– Simulate for the nominal values
𝑚 = 0.1, ℓ = 1, 𝑘0 = 0.02 and set 𝑎1 = 1;
– For ideal SMC:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 38
Sliding Mode Control
 Chattering
• Example: (Cont.)
1
– Consider a delayed control input by a transfer function
0.01𝑠+1 2

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 39
Sliding Mode Control
• Motivating Example
 Reduce Chattering
• Can we reduce the gain of SMC controller 𝑘?
– Take 𝑢 = 𝑢ො + 𝑣, where
෡ (𝑥)
𝑎1 𝑥2 +ℎ
𝑢ො = − ො
and 𝑣 = −𝛽(𝑥) sgn(𝑠)
𝑔(𝑥)
– This results in

– If the perturbation term 𝛿 𝑥 satisfies:

– Since 𝜚 is the upper bound of the perturbation, then it will be much

smaller than before. The switching gain is 𝛽 𝑥 = 𝜚 𝑥 + 𝛽0 .
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 40
Sliding Mode Control
 Reduce Chattering
• Example (Cont.)
– Reduced gain: Consider
ෝ = 0.125, ℓ෠ = 1, 𝑘෠ 0 = 0.025
𝑚 and set 𝑎1 = 1;

– Then

– Set SMC as:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 41
Sliding Mode Control
 Reduce Chattering
• Continuefy signum function

where

– Hence:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 42
Sliding Mode Control
 Reduce Chattering
• Continuous SMC
– Examine reaching phase by
1
𝑉 = 𝑠2 → 𝑉ሶ ≤ −𝜂 𝑠 when 𝑠 ≥𝜀
2
– That is outside the boundary layer 𝑠 ≥ 𝜀 .
– Therefore, if 𝑠(0) ≥ 𝜀, then 𝑠(𝑡) will be decreasing to sink in 𝑠 ≤ 𝜀 .
– This is reached in finite time, and remains in it forever.
– Inside the boundary layer where 𝑠 ≤ 𝜀, then 𝑥ሶ 1 = −𝑎1 𝑥1 + 𝑠.
1
– Consider a Lyapunov function 𝑉1 = 𝑥12 , then
2

– where, 0 < 𝜃 < 1. Thus the trajectory reaches to the set

𝜀
Ω𝜀 = 𝑥1 ≤ , 𝑠 ≤𝜀
𝑎1 𝜃1
UUB stable with ultimate bound reducible by 𝜀.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 43
Sliding Mode Control
 Reduce Chattering
• Continuous SMC Example: (Cont.)
– Inside the boundary layer: 𝑢 = −𝑘𝑠/𝜀, then

– Has a unique eq. point @ (𝑥ҧ1 , 0), where

– Approximate it by:
– One can show that (𝑥1ҧ , 0) is asymptotically stable.
– This means we have steady-state error of 𝑥1ҧ .
– Choosing 𝜀 is a trade off between the steady-state error and the control
input chattering.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 44
Sliding Mode Control
 Remove Steady-State Error
• Integral SMC: Example (Cont.)
– Let 𝑥0 = ∫ 𝑥1 𝑑𝑡, and write the augmented state-space:

– Choose 𝑠 = 𝑎0 𝑥0 + 𝑎1 𝑥1 + 𝑥2 (choose stable polynomials)

– If:

– Choose:

– To ensure:
– and finite reaching time
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 45
Sliding Mode Control
 Remove Steady-State Error
• Integral SMC: Example (Cont.)
– Inside the boundary layer:
– Similar to previous analysis:
• There is a unique eq. point @
– But because of integral action, the integral of the state has a steady
state error, and 𝜃 converges to the desired 𝛿1 .

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 46
Sliding Mode Control
• Single-Input SMC
 Consider an SI system as: 𝑥ሶ = 𝑓𝑔 𝑥 + 𝑔𝑔 𝑥 𝑢
• Use input-state feedback linearizing diffeomorphism 𝑧 = 𝑇(𝑥):
𝑧1ሶ = 𝑧2
𝑧2ሶ = 𝑧3
𝑛
⋮ ⇒ 𝑥 =𝑓 𝑥 +𝑏 𝑥 𝑢
𝑧𝑛−1
ሶ = 𝑧𝑛
𝜕𝑧𝑛 𝜕𝑧𝑛
𝑧𝑛ሶ = 𝑓 𝑥 + 𝑔 𝑥 𝑢
𝜕𝑥 𝑔 𝜕𝑥 𝑔
– In which,
𝜕𝑧𝑛
𝑓 𝑥 = 𝑓 𝑥 = 𝐿𝑛𝑓 𝑧1 = −𝛼(𝑥)/𝛽(𝑥)
𝜕𝑥 𝑔
𝜕𝑧𝑛
𝑏 𝑥 = 𝑔 𝑥 𝑢 = 𝐿𝑔 𝐿𝑛−1
𝑓 𝑧1 = 1/𝛽(𝑥)
𝜕𝑥 𝑔
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 47
Single-Input Sliding Mode Control
• In Nonlinear Systems
 Uncertainty is captured by norm-bounded functions.
𝑛
• Consider single-input system: 𝑥 =𝑓 𝑥 +𝑏 𝑥 𝑢
– Where 𝑓 𝑥 and 𝑏 𝑥 are unknown.
– Consider the nominal model by 𝑓መ 𝑥 and uncertainty bound by
𝑓መ − 𝑓 ≤ 𝐹 (unmatched or additive uncertainty)
– Furthermore, consider 0 < 𝑏𝑚𝑖𝑛 ≤ 𝑏 ≤ 𝑏𝑚𝑎𝑥 , the nominal model is
the geometric mean 𝑏෠ = 𝑏𝑚𝑖𝑛 𝑏𝑚𝑎𝑥 1/2 and the uncertainty
1/2
𝑏𝑚𝑎𝑥 𝑏෠
bound is 𝛽= by which 𝛽 −1 ≤ ≤ 𝛽 (matched or
𝑏𝑚𝑖𝑛 𝑏
multiplicative uncertainty).

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 48
Single-Input Sliding Mode Control
 Tracking Objective:
𝑛
• For the single-input system: 𝑥 =𝑓 𝑥 +𝑏 𝑥 𝑢
– The state vector is: 𝑥 = 𝑥 𝑥ሶ … 𝑥 𝑛−1 𝑇
– The system model has additive and multiplicative uncertainty
• The objective is:
– Tracking of the desired time varying state
𝑇
𝑥𝑑 = 𝑥𝑑 𝑥ሶ 𝑑 … 𝑥𝑑𝑛−1
– In presence of the modeling uncertainty on 𝑓, 𝑏.
• Define the tracking error:
𝑇
𝑥෤ = 𝑥 − 𝑥𝑑 = 𝑥෤ 𝑥෤ሶ … 𝑥෤ 𝑛−1
– Multi-state notion.
•
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 49
Single-Input Sliding Mode Control
• Sliding Manifold
 Define time varying sliding manifold 𝑆(𝑡)
• By a scalar equation 𝑠 ( 𝑥 ; 𝑡) = 0.
• For Example consider:
𝑑 𝑛−1
– 𝑠 𝑥; 𝑡 = +𝜆 𝑥෤ where 𝜆 > 0
𝑑𝑡

– For 𝑛 = 2: 𝑠 = 𝑥෤ሶ + 𝜆 𝑥෤ or for 𝑛 = 3: 𝑠 = 𝑥෤ሷ + 2𝜆𝑥෤ሶ + 𝜆2 𝑥෤

– For initial condition 𝑥𝑑 (0) = 𝑥(0) the problem of having perfect
tracking 𝑥 ≡ 𝑥𝑑 is equivalent to remain on the sliding manifold
𝑆 𝑡 ∀ 𝑡 > 0.
– The equation 𝑠 = 0 reveals a differential equation whose unique
solution is 𝑥෤ ≡ 0, given the matched initial condition.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 50
Single-Input Sliding Mode Control
• Sliding Manifold
• Converting multi-state tracking (in 𝑥) to a scalar stabilization
problem in 𝑠.
• For non matched initial condition 𝑥 (0) 𝑥𝑑 (0), we need
convergence to the sliding surface.
• Assume a Lyapunov Candidate:
1
𝑉 𝑠 = 𝑠2
2
– Suppose its derivative is hold negative definite through the control
input u:
1 𝑑 2
𝑉ሶ 𝑠 = 𝑠𝑠ሶ = 𝑠 ≤ −𝜂 𝑠 for 𝜂 > 0
2 𝑑𝑡

– This is called sliding condition.

– Geometrically: the squared distance to the surface (𝑠 2 ) decreased
monotonically along all trajectories.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 51
Single-Input Sliding Mode Control
• Sliding Manifold
 Graphical Representation:
• For n=2:

– Finite time reaching phase:

𝑉ሶ ≤ −𝜂 𝑠 Or 𝑉ሶ ≤ −𝜂 2𝑉 ⟹ 𝑠 𝑡 ≤ 𝑠 0 − 𝜂 𝑡
→ t 𝑟𝑒𝑎𝑐ℎ ≤ 𝑠 0 /𝜂
– 𝑠 𝑡 reaches to 𝑠 = 0 in finite time.
– On the surface 𝑠 = 0, the trajectory cannot leave it

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 52
Single-Input Sliding Mode Control
 Controller Design: VSC Law
• The control law consists of
𝑢 = 𝑢ො − 𝑘sgn(𝑠)
– In which 𝑢ො is a feedback linearizing controller, and

– The controller gain 𝑘 must be chosen such that the sliding

condition is satisfied for the uncertain system.

– Due to the switching nature of signum function, the controller has

a variable structure

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 53
Single-Input Sliding Mode Control
• Controller Design
 Example 1: (Basic System)
• Consider the uncertain 2nd order system: 𝑥ሷ = 𝑓 + 𝑢
– In which the function 𝑓 can be captured by norm-bounded
uncertainty (additive or non-matched)
𝑓መ − 𝑓 ≤ 𝐹
– For example consider:
with
– Set:
Then:
𝑑
– Consider Sliding manifold: 𝑠 𝑥 = + 𝜆 𝑥෤ = 𝑥෤ሶ + 𝜆 𝑥෤
𝑑𝑡

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 54
Single-Input Sliding Mode Control
• Controller Design
 Example 1: (Cont.)
• Satisfy Sliding Condition:

– Feedback linearizing controller:

– Variable structure control:
– Check sliding condition

– Satisfy sliding condition

1 d 2
V (s)  s  s  s   s  0
2 dt
– Hence, if
1 d 2
F  k    k  F   s   s
2 dt
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 55
Single-Input Sliding Mode Control
• Controller Design
 Example 2: (Integral Control)
• For the previous example, assume the integral of the tracking
error must converge to zero.
– Consider the following sliding surface:

– Feedback linearizing controller:

– Variable structure control:

1 d 2
– Sliding condition k  F   s   s
2 dt
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 56
Single-Input Sliding Mode Control
• Controller Design
 Example 3: (Full Uncertainty)
• Consider the system:
– In which the function 𝑓 can be captured by norm-bounded uncertainty
(additive or non-matched)
𝑓መ − 𝑓 ≤ 𝐹
– And the function 𝑏 can be captured by norm-bounded multiplicative
uncertainty (matched)
– consider 0 < 𝑏𝑚𝑖𝑛 ≤ 𝑏 ≤ 𝑏𝑚𝑎𝑥 , the nominal model is the geometric
1/2
𝑏𝑚𝑎𝑥
mean 𝑏෠ = 𝑏𝑚𝑖𝑛 𝑏𝑚𝑎𝑥 1/2
and the uncertainty bound is 𝛽 =
𝑏𝑚𝑖𝑛
𝑏෠
by which 𝛽 −1 ≤ ≤𝛽
𝑏
𝑑
– Consider Sliding Manifold: 𝑠 𝑥 = + 𝜆 𝑥෤ = 𝑥෤ሶ + 𝜆 𝑥෤
𝑑𝑡
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 57
Single-Input Sliding Mode Control
• Controller Design
 Example 3: (cont.)
1 d 2
• Satisfy Sliding Condition: V ( x)  s  s  s   s  0
2 dt
– Find 𝑠:ሶ
s  x   x  f  bu  xd   x

– Feedback linearizing controller: uˆ  fˆ  x d   x

– Variable structure controller: u  bˆ 1 uˆ  k sgn(s ) 

– Check sliding condition
s  f  bbˆ 1 uˆ  k sgn( s )   xd   x
 ( f  bbˆ 1 fˆ )  (1  bbˆ 1 )( xd   x)  bbˆ 1k sgn( s )

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 58
Single-Input Sliding Mode Control
• Controller Design
 Example 3: (cont.)
– Satisfy sliding condition
s  s  ( f  bbˆ 1 fˆ )  (1  bbˆ 1 )( xd   x)  s  bbˆ 1k s   s
 
– Hence, ˆ 1 ( f  bbˆ 1 fˆ )  (1  bbˆ 1 )( x   x)   
k  bb
 d 

– Since: f  fˆ  F and ˆ 1  
bb

– Then k   ( f  fˆ )  ( fˆ  bbˆ 1 fˆ )  (1   )( xd   x)  

k   F    (   1) fˆ  x d   x
– Hence:
k   (F   )  (   1) uˆ
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 59
Single-Input Sliding Mode Control
• Continuous Sliding Controller
 Chattering
• Discontinuous Control across s(t)
• Imperfect Control
– Discrete time OR Quantization error
– Chattering: Oscillation across s(t)

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 60
Single-Input Sliding Mode Control
• Continuous Sliding Mode
 Smooth out control discontinuity
• Boundary Layer
– Linear Interpolation

– Tangent hyperbolic

– Fuzzy Switching
– ...
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 61
Single-Input Sliding Mode Control
• Continuous Sliding Mode
 Boundary Layer: Linear Interpolation x
𝑠 𝜑
• Change: sgn 𝑠 ⇒ 𝑠𝑎𝑡
𝜙
• Outside Boundary Layer
– Controller Law as before
– Satisfy sliding condition
– Attractive Boundary Layer ε x
• Inside Boundary Layer 𝑠 = 0
– Linear interpolation
– Invariant Boundary Layer:
Trajectories starting inside B.L.
remain inside B.L.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 62
Single-Input Sliding Mode Control
• Continuous Sliding Mode
 Boundary Layer: Linear Interpolation
• Perfect tracking UUB with precision ε
x
𝜑

– Boundary Layer Thickness: 𝜑

– Boundary Layer Width: 𝜀 = 𝜑/𝜆𝑛−1
t  0, x (i ) (t )  (2 )i  i  0, , n 1 ε x
𝑠 = 0

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 63
Single-Input Sliding Mode Control
• Continuous Sliding Mode
 Boundary Layer: Linear Interpolation
• Example: (Basic System)
– Reconsider: with
– Set Then:
– For and 𝜆 = 20 , 𝜂 = 0.1
• Switching control law:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 64
Single-Input Sliding Mode Control
• Continuous Sliding Mode
 Boundary Layer: Linear Interpolation
• Example 1: (Basic System cont.)
• Smoothed control law:
– For and 𝜆 = 20 , 𝜂 = 0.1, 𝜑 = 0.1

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 65
Single-Input Sliding Mode Control
• Robust Input-Output Feedback Linearization Tracking
 Remind Tracking Control in IOFL
(𝜌−1) 𝑇 (𝜌−1) 𝑇
• For 𝜇 = 𝑦 𝑦ሶ … 𝑦 and 𝜇𝑑 = 𝑦𝑑 𝑦ሶ 𝑑 … 𝑦𝑑
• And tracking error 𝜇෤ 𝑡 = 𝜇 𝑡 − 𝜇𝑑 (𝑡)
• Theorem 13.2
– Input-output feedback linearization control:
1 𝜌 (𝜌) 𝜌−1
𝑢ො 𝑥 = 𝜌−1 [−𝐿𝑓 𝑦 + 𝑦𝑑 − 𝑘𝜌−1 𝑦 − ⋯ − 𝑘1 𝑦ሶ − 𝑘0 𝑦]
𝐿𝑔 𝐿𝑓 𝜇1
– Robustify the controller by 𝑢 = 𝑢ො 𝑥 + 𝑣 where 𝑣 = −𝛽 𝑥 sgn 𝑠
– By sliding manifold
𝑠 = 𝑘0 𝜇෤1 + 𝑘1 𝜇෤2 + ⋯ + 𝑘𝜌−1 𝜇෤𝜌 = 0
– Then
𝜌−1
𝑠ሶ = 𝐿𝑔 𝐿𝑓 ℎ 𝑥 𝑣 + Δ(𝑡, 𝑥, 𝑣)
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 66
Single-Input Sliding Mode Control
• Robust Input-Output Feedback Linearization
• For bounded uncertainty

– where 𝜚 and 𝜅_0 are known

• Then the switching controller 𝑣 = −𝛽 𝑥 sgn 𝑠

– where
𝛽 𝑥 ≥ 𝜚(𝑥)/(1 − 𝜅0 ) + 𝛽0 for 𝛽0 > 0

– Will result into robust tracking of the desired output 𝑦𝑑 (𝑡) in presence
of uncertainties.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 67
Contents

Backstepping
1 Design idea, integrator and general backstepping, recursive backstepping, examples.

Sliding Mode Controller

2
Motivating example, chattering, continuous SMS, integral SMC. Single-input SMC; tracking
objective, sliding manifold, variable structure control, continuous SMC, robust IOFL. Multi-Input
SMC; regular form, sliding condition, chattering reduction, UUB stability, unmatched uncertainty.

Lyapunov Redesign
3 Robust stabilization, nominal Lyapunov function, 2-norm and ∞-norm perturbation, boundary
layer, nonlinear damping, examples.

In this chapter we review a few nonlinear control Lyapunov-based design methods. First backstepping method is introduced to iteratively
generate a complex Lyapunov function from low order subsystems. Sliding mode control, as one of the most celebrated robust control schemes
are reviewed for single-input and multi-input systems. Finally, Lyapunov redesign is described, on how a robust stabilizing controller can be
generated by robustifying a nominal nonlinear controller.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 68
Sliding Mode Control
• Generalize to Multi-Input SMC
 Stabilization
• Consider the system

– Where 𝑥 ∈ 𝑅𝑛 , 𝑢 ∈ 𝑅𝑝 and 𝑓, 𝐵, 𝐺, 𝐸 are smooth functions in 𝐷 ⊂ 𝑅𝑛 .

– 𝐸 is known and nonsingular
– 𝐺 and 𝛿 are uncertain.
– 𝐺 𝑥 is diagonal matrix with elements 𝑔𝑖 𝑥 ≥ 𝑔0 > 0
– 𝑓𝑎 0,0 = 0, 𝑓𝑏 0,0 = 0
• Control objective: stabilize the origin ∀ uncertainties in 𝐺, 𝛿.
– Use the diffeomorphism 𝑇: 𝐷 → 𝑅𝑛 such that

Where 𝐼 is the 𝑝 × 𝑝 identity matrix.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 69
Multi-Input Sliding Mode Control
• Regular Form
 The system is transformed to:
𝜂ሶ = 𝑓𝑎 𝜂, 𝜉
𝜉ሶ = 𝑓𝑏 𝜂, 𝜉 + 𝐺 𝑥 𝐸 𝑥 𝑢 + 𝛿(𝑡, 𝑥, 𝑢)
𝜂
𝑥 = 𝜉 ∈ 𝑅𝑛 , 𝜂 ∈ 𝑅𝑛−𝑝 , 𝜉, 𝑢 ∈ 𝑅𝑝

• Sliding Manifold: Design 𝑠 = 𝜉 − 𝜙 𝜂 = 0

– When motion is restricted to this manifold, then
𝜂ሶ = 𝑓𝑎 (𝜂, 𝜙 𝜂 ) has an asy. Stable eq. point @ origin.

– Design of 𝜙 𝜂 is to solve stabilization problem for 𝜂ሶ = 𝑓𝑎 𝜂, 𝜉

• Feedback linearization or
• Backstepping, …
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 70
Multi-Input Sliding Mode Control
• Regular Form
 Sliding Condition
• Derive 𝑠ሶ as:
• Design Control Law:
– Simple and pure switching or
– Nominal plus switching law:

– where 𝐿 𝑥 = 𝐺෠ −1 𝑥 .
– In absence of uncertainty: 𝐺 = 𝐺,෠ 𝛿 = 0, this results in 𝑠ሶ = 0.
– In presence of uncertainty, the known dynamics is cancelled and

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 71
Multi-Input Sliding Mode Control
 Sliding Condition
• where Δ𝑖 is the 𝑖 th component of

• and 𝑔𝑖 is the 𝑖 th diagonal element of 𝐺.

– Assume bound for uncertainty

– where, 𝜚 𝑥 ≥ 0, and 𝜅0 ∈ [0,1) are the known bounds.

• Design switching 𝑣 as: (𝛽0 > 0)

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 72
Multi-Input Sliding Mode Control
 Sliding Condition
• By this means
1
– For all 𝑝 Lyapunov function 𝑉𝑖 = 𝑠𝑖2 ,
2

– and

– This results in finite reaching time to 𝑠 = 0.

• Continuous approximation:
– Replace signum by saturation function:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 73
Multi-Input Sliding Mode Control
 Chattering Reduction
• Continuous Approximation:
1
– For all 𝑝 Lyapunov function 𝑉𝑖 = 𝑠𝑖2 ,
2

– In the region 𝑠𝑖 ≤ 𝜀.

– 𝑠𝑖 𝑡 converges to the boundary layer set: |𝑠𝑖 | ≤ 𝜀, 1 ≤ 𝑖 ≤ 𝑝

• UUB stability
– Assume for the sliding manifold 𝜉 = 𝜙 𝜂 → ∃ 𝑉 𝜂 ∋

– Where 𝛼1 , 𝛼2 , 𝛼3 , 𝛾 are class K functions

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 74
Multi-Input Sliding Mode Control
 Chattering Reduction
• UUB Stability
– Note that

– For some positive constant 𝑘1 , set a class K function 𝛼 by

– Then

– One can show that the trajectories will reach to positively invariant
set:

– This set can be made arbitrary small by choosing 𝜀 small enough.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 75
Multi-Input Sliding Mode Control
 Chattering Reduction
• UUB Stability

– If 𝛿 vanishes at origin, asymptotic stability can be achieved.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 76
Multi-Input Sliding Mode Control
 Design Summary:
• Sliding Manifold: 𝑠 = 𝜉 − 𝜙 𝜂 = 0
– Design 𝜉 = 𝜙 𝜂 to stabilize 𝜂ሶ = 𝑓𝑎 (𝜂, 𝜙 𝜂 )
• Control Law: 𝑢
– Design simplified control law 𝑢 = 𝐸 −1 𝑣 or
–
• Uncertainty Bound
– Estimate 𝜚(𝑥) and 𝜅0 in (14.10) depending on your choice of 𝑢
• Design switching 𝑣 as to satisfy uncertainty bound

• Or the continuous controller to reduce chattering

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 77
Multi-Input Sliding Mode Control
 Unmatched Uncertainty
• Consider this equation instead of 14.1

• Use 𝑇(𝑥) as before to transform it into:

– where

– The term 𝛿𝑏 is added due to unmatched uncertainty

– It changes the upper bound on Δ𝑖 /𝑔𝑖 .
– The term 𝛿𝑎 changes the reduced order model to

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 78
Multi-Input Sliding Mode Control
 Unmatched Uncertainty
• Design 𝜙 𝜂
– To ensure asymptotic stability of the origin 𝜂 = 0 in presence of
uncertainty.
– This needs a robust stabilization technique,
• High gain feedback or
• Lyapunov redesign
• The main difference in unmatched uncertainty
– In matched uncertainty, if the upper bound is known
• SMC guarantees robustness
– In unmatched uncertainty, such guarantee is not available.
• Restrict the size of uncertainty to robustly stabilize
– See the next two examples
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 79
Multi-Input Sliding Mode Control
• Example 1:
 Global Stabilization
• Consider the system:
𝑥ሶ1 = 𝑥2 + 𝜃1 𝑥1 sin 𝑥2
𝑥ሶ 2 = 𝜃2 𝑥22 + 𝑥1 + 𝑢 𝜃1 ≤ 𝑎 and 𝜃2 ≤ 𝑏
– The system is in regular form with 𝜂 = 𝑥1 and 𝜉 = 𝑥2 .
– The uncertainty 𝜃2 is matched, while that for 𝜃1 is unmatched.
• Start with the first equation:
– Design 𝑥2 to robustly stabilize 𝑥1 = 0.
– Use high gain controller, 𝑥2 = −𝑘𝑥1 , 𝑘 > 𝑎, then

– For the sliding manifold 𝑠 = 𝑥2 + 𝑘𝑥1 = 0,

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 80
Multi-Input Sliding Mode Control
 Example 1: (Cont.)
• To cancel the nonlinear terms:
– Choose: 𝑢 = −𝑥1 − 𝑘𝑥2 + 𝑣
– Then: 𝑠ሶ = 𝑣 + Δ 𝑥
where, Δ 𝑥 = 𝜃2 𝑥22 + 𝑘𝜃1 𝑥1 sin 𝑥2
– Since Δ 𝑥 ≤ 𝑎𝑘 𝑥1 + 𝑏𝑥22
– Design high gain switching controller as
𝛽 𝑥 = 𝑎𝑘 𝑥1 + 𝑏𝑥22 + 𝛽0 , 𝛽0 > 0
and 𝑢 = −𝑥1 − 𝑘𝑥2 − 𝛽 𝑥 sgn(𝑠)
– This controller or its continuous approximation with small 𝜀,
globally stabilizes the origin.
– Note that we could robustly stabilize the origin for unmatched
uncertainty 𝜃1 ≤ 𝑎 without having to restrict 𝑎.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 81
Multi-Input Sliding Mode Control
• Example 2:
 limited Stabilization
• Consider the system:
𝑥ሶ1 = 𝑥1 + (1 − 𝜃1 )𝑥2
𝑥ሶ 2 = 𝜃2 𝑥22 + 𝑥1 + 𝑢 𝜃1 ≤ 𝑎 and 𝜃2 ≤ 𝑏
• Start with the first equation:
– Design 𝑥2 to robustly stabilize 𝑥1 = 0.
– Note that we loose controllability if 𝜃1 = 1.
• Limit 𝜃1 = 1 or 𝑎 < 1 to stabilize the origin
– As before use 𝑥2 = −𝑘𝑥1 , then

– For , 𝑘 > 1/(1 − 𝑎), the origin can be robustly stabilized.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 82
Multi-Input Sliding Mode Control
• Example 2: (Cont.)
 limited Stabilization
• Proceeding like example 1:
– Design SMC controller:

– where

– Unlike the previous example, for this unmatched uncertainty,

robust stabilization may be achieved for a limited uncertainty
bound 𝜃1 ≤ 𝑎 < 1.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 83
Multi-Input Sliding Mode Control
• Example 3:
 UUB Bound Estimation
• Consider the system:
– pendulum with horizontal acceleration ℎ(𝑡)
𝑇
𝑚𝑙 𝜃ሷ + 𝑚𝑔 sin 𝜃 + 𝑘𝑙𝜃ሶ = + 𝑚ℎ 𝑡 cos 𝜃
𝑙
– Where 0.9 ≤ 𝑙 ≤ 1.1, 0.5 ≤ 𝑚 ≤ 1.5, 0 ≤ 𝑘 ≤ 0.2, ℎ 𝑡 ≤ 1, 𝑔 = 9.8
• Control Objective:
– Design SMC to achieve ultimate bounds 𝜃 ≤ 0.01 and 𝜃ሶ ≤ 0.01
• State-space representation
𝑥ሶ1 = 𝑥2
ቊ
𝑥ሶ 2 = −𝑎 sin 𝑥1 − 𝑏𝑥2 + 𝑐𝑢 + 𝜙 𝑡 cos 𝑥1
1
– where 𝑥1 = 𝜃 , 𝑥2 = 𝜃ሶ , 𝑢 = 𝑇, 𝑎 = 𝑔/𝑙, 𝑏 = 𝑘/𝑚, 𝑐 = , 𝜑 𝑡 = ℎ(𝑡)/𝑙
𝑚𝑙 2
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 84
Multi-Input Sliding Mode Control
• Example 3: (Cont.)
• Consider the Sliding manifold: 𝑠 = 𝑥1 + 𝑥2
– Then 𝑠ሶ = 𝑥2 − asin 𝑥1 − 𝑏𝑥2 + 𝑐𝑢 + 𝜙 𝑡 cos 𝑥1 = 𝑐 𝑢 + 𝛿
1
– While, 𝛿 = 𝑥2 − 𝑎𝑠𝑖𝑛 𝑥1 − 𝑏𝑥2 + 𝑐𝑢 + 𝜙 𝑡 cos 𝑥1
𝑐
𝑎 1−𝑏 𝜙 𝑡
𝛿 ≤ 𝑥1 + 𝑥2 + ≤ 16.1865 𝑥1 + 1.815 𝑥2 + 1.1111
𝑐 𝑐 𝑐
• Controller Design:
𝑠
𝑢 = − 16.1865 𝑥1 + 1.815 𝑥2 + 2 𝑠𝑎𝑡
𝜀
– The trajectory reaches the boundary layer { 𝑠 ≤ 𝜀} in finite time.
• Inside the boundary layer
1
𝑥ሶ 1 = −𝑥1 + 𝑠, 𝑉 = 𝑥12
2
𝜀
𝑉1ሶ = −𝑥12 + 𝑥1 𝑠 ≤ −𝑥12 + 𝑥1 𝜀 ≤ − 1 − 𝛾 𝑥12 , ∀ 𝑥1 ≥
𝛾
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 85
Multi-Input Sliding Mode Control
• Example 3: (Cont.)
• The trajectory reaches the set
𝜀
– Ω𝜀 = 𝑥1 ≤ , 𝑥1 + 𝑥2 ≤ 𝜀 in finite time.
𝛾
• Furthermore
1
𝑥2 = 𝑥1 + 𝑥2 − 𝑥1 ≤ 𝑥1 + 𝑥2 + 𝑥1 ≤ 1 + 𝜀
𝛾
– As an example, for 𝛾 = 0.9
1
𝑥2 ≤ 1 + 𝜀 ≤ 2.11 𝜀
𝛾
– And to assure that 𝜃ሶ ≤ 0.01:
0.01
𝑥2 ≤ 0.01 → 𝜀 ≥ = 4.7 × 10−3
2.11
– By this means it is assured that 𝜃 ≤ 0.01.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 86
Contents

Backstepping
1 Design idea, integrator and general backstepping, recursive backstepping, examples.

Sliding Mode Controller

2
Motivating example, chattering, continuous SMS, integral SMC. Single-input SMC; tracking
objective, sliding manifold, variable structure control, continuous SMC, robust IOFL. Multi-Input
SMC; regular form, sliding condition, chattering reduction, UUB stability, unmatched uncertainty.

Lyapunov Redesign
3 Robust stabilization, nominal Lyapunov function, 2-norm and ∞-norm perturbation, boundary
layer, nonlinear damping, examples.

In this chapter we review a few nonlinear control Lyapunov-based design methods. First backstepping method is introduced to iteratively
generate a complex Lyapunov function from low order subsystems. Sliding mode control, as one of the most celebrated robust control schemes
are reviewed for single-input and multi-input systems. Finally, Lyapunov redesign is described, on how a robust stabilizing controller can be
generated by robustifying a nominal nonlinear controller.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 87
Lyapunov Redesign
• Robust Stabilization
 Consider the uncertain system:

 In which
• 𝑥 ∈ 𝑅𝑛 , 𝑢 ∈ 𝑅𝑝 ,
• The functions 𝑓 and 𝐺 are known (nominal model)
• The function δ is the modeling uncertainty.
– It enters the equation as 𝑢 and satisfies matching condition.
• Nominal model

– Suppose we found a stabilizing controller 𝑢 = 𝜓 𝑡, 𝑥  the origin of

– is uniformly asymptotically stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 88
Lyapunov Redesign
• Robust Stabilization
• Consider  a Lyapunov Function 𝑉 (𝑡 , 𝑥) 

– Where 𝛼1 , 𝛼2 , 𝛼3 are class K functions.

– Assume that with 𝑢 = 𝜓 𝑡, 𝑥 + 𝑣, 𝛿 satisfies:

– This bound is the only thing that shall be known for the uncertainty.
– Lyapunov Redesign is to redesign 𝑢 = 𝜓 𝑡, 𝑥 + 𝛾 𝑡, 𝑥 , ∋ the
uncertain system is robustly stabilized.
– Apply controller 𝑢 = 𝜓 𝑡, 𝑥 + 𝑣 to the uncertain system:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 89
Lyapunov Redesign
• Robust Stabilization
• This is a perturbed version of the closed-loop dynamics
– Recalculate the derivative of the Lyapunov function

– Set 𝑤 𝑇 = [𝜕𝑉/𝜕𝑥]𝐺, hence:

𝑉ሶ ≤ −𝛼3 𝑥 + 𝑤 𝑇 𝑣 + 𝑤 𝑇 𝛿
• Due to matching condition 𝑣 enters the same place as 𝛿
• Choose 𝑣 to cancel the effect of uncertainty 𝛿
– Suppose
– Hence

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 90
Lyapunov Redesign
• Robust Stabilization
• Robust Controller 1st Choice:

– where 𝜂 𝑡, 𝑥 ≥ 𝜌 𝑡, 𝑥 ∀(𝑡, 𝑥), then

– Choosing 𝜂 𝑡, 𝑥 ≥ 𝜌 𝑡, 𝑥 /(1 − 𝜅0 ) in the domain 𝐷, yields

ሶ 𝑥) becomes negative definite.

– Hence 𝑉(𝑡,

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 91
Lyapunov Redesign
• Robust Stabilization
• Robust Controller 2nd Choice:
– Suppose infinity norm bound of (14.35)

– Then
– Choose
– where 𝜂 𝑡, 𝑥 ≥ 𝜌 𝑡, 𝑥 /(1 − 𝜅0 ) in the domain 𝐷, yields

ሶ 𝑥) becomes negative definite.

– Hence 𝑉(𝑡,
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 92
Lyapunov Redesign
• Robust Stabilization
 Boundary Layer
• Approximate by a continuous control law:
– 1st choice:

– Hence:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 93
Lyapunov Redesign
 Boundary Layer
– Then the term
– Attain its maximum value 𝜀/4, at 𝜂 𝑤 2 = 𝜀/2.
– Therefore,

– is satisfied, irrespective to the value of 𝜂 𝑤 2.

– Take 𝑟 > 0 such that 𝐵𝑟 ⊂ 𝐷, and
– Choose 𝜀 < 2𝛼3 𝛼2−1 𝛼1 𝑟 /(1 − 𝜅0 )
– Set 𝜇 = 𝛼3−1 (𝜀(1 − 𝜅0 )/2) < 𝛼2−1 𝛼1 𝑟 .
– Then

– Hence, The solutions of the closed-loop systems are uniformly

ultimately bounded by a class K function of 𝜀.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 94
Lyapunov Redesign
• Robust Stabilization

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 95
Lyapunov Redesign
• Robust Stabilization
• Notes:
– The discontinuous Lyapunov redesign stabilize the origin
asymptotically.
– The continuous version guarantees UUB stability.
– The ultimate bound 𝑏(𝜀) can be set arbitrarily small by choosing
small 𝜀.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 96
Lyapunov Redesign
• Robust Stabilization
 Example 1: (Robust Input-State Feedback linearization)
• Consider the input-state feedback linearizable system

– With diffeomorphism 𝑇(𝑥) transformed into normal form

• Consider the perturbed system as:

– which satisfies matching condition

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 97
Lyapunov Redesign
• Robust Stabilization
 Example 1: (Cont.)
• Rewrite perturbed system in 𝑧:

– where 𝛿 𝑥, 𝑢 = Δ1 𝑥 + Δ2 𝑥 𝑢. Linearize the nominal system by

– Where 𝐾 is chosen  (𝐴 − 𝐵𝐾) is Hurwitz.

• For the F.L. nominal system: 𝑧ሶ = (𝐴 − 𝐵𝐾) 𝑧
– The Lyapunov function 𝑉 𝑧 = 𝑧 𝑇 𝑃𝑧, where 𝑃 satisfies:

– With 𝑢 = 𝜓 𝑥 + 𝑣 the uncertain function 𝛿(𝑥, 𝑢) satisfies:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 98
Lyapunov Redesign
 Example 1: (Cont.)
– Thus to satisfy (14.35) we need

– and

• The 1st inequality is restrictive, since it puts a definite limit on

perturbation 𝛿2 .
• The 2nd inequality is not restrictive, since we don’t need to have a small
𝜌, we just need to have an estimate of 𝜌.
• Controller Design
– Consider two-norms for the uncertainties:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 99
Lyapunov Redesign
 Example 1: (Cont.)
• Controller Design
– Use controller

– with

– All assumptions of theorem 14.3 is satisfied, with

– Thus, with the feedback control 𝑢 = 𝜓 𝑥 + 𝑣, the origin of the

perturbed closed-loop system is exponentially stable.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 100
Lyapunov Redesign
• Robust Stabilization
 Example 2: Inverted Pendulum
• Consider the pendulum equation with 𝛿1 = 𝜋:

– The I-S F.L. controller to stabilize the nominal system is:

– Where the hatted parameters are the nominal values, and 𝑘𝑖 ’𝑠 are
0 1
selected such that 𝐴 + 𝐵𝐾 = is Hurwitz.
𝑘1 𝑘2 − 𝑏
– With 𝑢 = 𝜓 𝑥 + 𝑣, the uncertainty term 𝛿 is given by:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 101
Lyapunov Redesign
• Robust Stabilization
 Example 2: (Cont.)
• Hence:
– Where:

– Assuming 𝜅0 < 1 , the controller 𝑢 = 𝜓 𝑥 + 𝑣, the origin

becomes globally UUB stable.
– Note the contribution of the additional term to cope with the
modeling uncertainty.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 102
Lyapunov Redesign
• Robust Stabilization
 Example 3:
• Consider the system:
– Pendulum with horizontal acceleration ℎ(𝑡), neglecting friction 𝑏.
𝑥ሶ1 = 𝑥2
ቊ
𝑥ሶ 2 = −𝑎 sin 𝑥1 + 𝑐𝑢 + ℎ 𝑡 cos 𝑥1
– Assume ℎ 𝑡 ≤ 𝐻 for all 𝑡 ≥ 0.
– The controller and the uncertainty bounds 𝜌1 , 𝜅0 are as in Example 2.
– The uncertain term 𝛿 satisfies:
– This time 𝜌 𝑥 = 𝜌1 𝑥 2 + 𝐻/𝑐,Ƹ hence does not vanish @ 𝑥 = 0.
– The choice of 𝜂 in the control law (14.41) shall satisfy
𝜂 ≥ (𝜌1 𝑥 2 + 𝐻/𝑐)/(1
Ƹ − 𝜅0 )
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 103
Lyapunov Redesign
• Robust Stabilization
 Example 3: (Cont.)
• Choose:
𝜂 𝑥 = 𝜂0 + 𝜌1 𝑥 2 /(1 − 𝜅0 ) with 𝜂0 ≥ 𝐻/𝑐(1
Ƹ − 𝜅0 )

– This is different from the previous example for vanishing disturbance,

where 𝜂(0) was equal to 1.
– In here 𝜂 0 = 𝜂0 ≥ 𝐻/𝑐(1 Ƹ − 𝜅0 ) to remove nonvanishing disturbance
term ℎ 𝑡 cos 𝑥1 .

• By this controller:
– The solutions of the closed-loop system are UUB with the ultimate
bound proportional to 𝜀.
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 104
Lyapunov Redesign
• Nonlinear Damping
 Unknown Uncertainty Upper Bound:
• Reconsider (14.30) with 𝛿 𝑡, 𝑥, 𝑢 = Γ 𝑡, 𝑥 𝛿0 (𝑡, 𝑥, 𝑢)

– Consider the Lyapunov function and conditions (14.33) and (14.34).

– Then

– Set 𝑤 𝑇 = [𝜕𝑉/𝜕𝑥]𝐺, and take:

– Then

– In which 𝑘0 is an unknown upper bound on 𝛿0 .

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 105
Lyapunov Redesign
• Nonlinear Damping
– The term
– attains a maximum value 𝑘02 /4𝑘 at 𝑤 2 Γ 2 = 𝑘0 /2𝑘
– Therefore,

– Since 𝛼3 is a class K∞, therefore 𝑉ሶ is negative outside some ball.

– Uniformly bounded solution of the closed-loop system is given in:

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 106
Lyapunov Redesign
• Nonlinear Damping
 Example
• Consider the scalar system 𝑥ሶ = 𝑥 2 + 𝑢 + 𝑥𝛿0
– Where, 𝛿0 (𝑡) is bounded.
– With nominal stabilizing control 𝜓 𝑥 = −𝑥 2 − 𝑥
– The Lyapunov function 𝑉 𝑥 = 𝑥 2 satisfies (14.33) and (14.34) globally
for 𝛼1 𝑟 = 𝛼2 𝑟 = 𝛼3 𝑟 = 𝑟 2 .
– The nonlinear damping term (14.48) with 𝑘 = 1, is given by 𝑣 = −2𝑥 3 .
– The closed-loop system
𝑥ሶ = −𝑥 − 2𝑥 3 + 𝑥𝛿0 𝑡
– Has a bounded solution for any bounded 𝛿0 𝑡 .
– This is due to the nonlinear damping −2𝑥 3 .
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 Nonlinear Control
Chapter 7: Nonlinear Controller
Design
To read more and see the course videos
visit our course website:
http://aras.kntu.ac.ir/arascourses/nonlinear-control/

Thank You
Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,
Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020
 About Hamid D. Taghirad
Hamid D. Taghirad has received his B.Sc. degree in mechanical engineering
from Sharif University of Technology, Tehran, Iran, in 1989, his M.Sc. in mechanical
engineering in 1993, and his Ph.D. in electrical engineering in 1997, both
from McGill University, Montreal, Canada. He is currently the University Vice-
Chancellor for Global strategies and International Affairs, Professor and the Director
of the Advanced Robotics and Automated System (ARAS), Department of Systems
and Control, Faculty of Electrical Engineering, K. N. Toosi University of Technology,
Hamid D. Taghirad Tehran, Iran. He is a senior member of IEEE, and Editorial board of International
Journal of Robotics: Theory and Application, and International Journal of Advanced
Professor Robotic Systems. His research interest is robust and nonlinear control applied to
robotic systems. His publications include five books, and more than 250 papers in
international Journals and conference proceedings.

Nonlinear Control K. N. Toosi University of Technology, Faculty of Electrical Engineering,

Prof. Hamid D. Taghirad Department of Systems and Control, Advanced Robotics and Automated Systems October 21, 2020