Nonlinear Control

Chapter 6: Feedback Linearization

In this chapter we review feedback linearization methods, as
the most fundamental technique in the design of state
feedback control for nonlinear systems. First a quick review on
differential geometry is presented. Then a generalized methods
for input-output and input-state feedback linearization.

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 [Link] 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
Introduction
1 Motivation examples on input-output, and input-state
feedback linearization, internal and zero dynamics.

Differential Geometry
2 Gradient and Jacobian of vector fields, Lie Derivative and Lie
bracket, involutivity, and diffeomorphism.

Input-Output Feedback Linearization

3 Relative degree, normal form, zero dynamics, non-minimum
phase systems, state feedback control, regulation and tracking.

Input-State Feedback Linearization

4 Feasibility conditions, suitable diffeomorphism, state-feedback
control, regulation and tracking.

In this chapter we review feedback linearization methods, as the most fundamental technique in the design of state feedback control for
nonlinear systems. The required mathematics and conditions are reviewed through a quick introduction to differential geometry. Then a
generalized methods for input-output and input-state feedback linearization.
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
Feedback Linearization
• Introduction
 Consider the nonlinear system
𝑥ሶ = 𝑓 𝑥 + 𝐺 𝑥 𝑢
𝑦=ℎ 𝑥
• Pose a question whether ∃ a state-feedback control
𝑢 =𝛼 𝑥 +𝛽 𝑥 𝑣
• And the state transformation
𝑧 = 𝑇(𝑥)
• That transform the nonlinear system through feedback to a
linear system of the form
𝑧ሶ = 𝐴𝑧 + 𝐵𝑣
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
Feedback Linearization
• Motivation Input-State Feedback linearization
 Example 1: Consider the system

• By inspection choose 𝑢 as
• To cancel the nonlinear term, and yield to
• For stabilization use a state-feedback
control
• The closed loop system is fully linear and
stable:
• The nonlinear control input 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
 7
Feedback Linearization
• Motivation Input-State Feedback linearization
 Example 2: Consider the system:

• We cannot simply cancel the nonlinear term

• Use the state transformation
• The system in 𝑧 coordinate:
• Now choose u such that to cancel the
nonlinear term
𝜋 𝜋
• This is well defined in − < 𝑥2 <
2 2
• Use inverse of transformation to go to the 𝑥
coordinates,
• which is well defined in −𝑎 < 𝑧 < 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
 8
Feedback Linearization
• Motivation Input-State Feedback Linearization
• Block diagram

• Characteristics
– The result is working in a large region, but it is not global.
– The linearization is done through feedback, this is fundamentally
different from Lyapunov linearization.
– The controller depends on the system model and needs state
measurements.
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
Feedback Linearization
• Motivation Input-Output Feedback linearization
 Example 2: (Cont.)
• Consider a desirable output to be controlled: i.e. 𝑦 = 𝑥2 .
• By the change of variable and control law as before:

• We have:

• While the state equation is linear by the output 𝑦 is not!

• By inspection use 𝑢 = 𝑥12 + 𝑣 to linearize both the state and
the output:

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
Feedback Linearization
• Anything Missing?
 What about 𝑥1 ?
• This dynamics is not controlled: 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
 11
Feedback Linearization
• Motivation Input-Output Feedback linearization
 Example 3:
• System Dynamics:

– To generate a direct relation between 𝑦 and 𝑢, differentiate 𝑦:

– Continue

– 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
 12
Feedback Linearization
• Motivation Input-Output Feedback linearization
 Example 3: (Cont.)
– By inspection design the control law as:

– Then the input-output map is linearized.

– Design the linear part of the controller:

– The error dynamics will be linear and 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
Feedback Linearization
• Motivation Input-Output Feedback linearization
 Example 3: (Cont.)
• Note that
– System is of order 3, the error dynamics is of order 2!
– A first order dynamics is not seen, though not controlled.
– This is called internal dynamics.
– For this example the internal dynamics is:

– Which is stable, hence, the controller would work well for the output,
without any worries about the last state becoming unstable.
• This could be analyzed by Zero Dynamics:
– The zero dynamics is the internal dynamics of the system when the
system output is kept zero by the input u.
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
Feedback Linearization
• Motivation Input-Output Feedback linearization
 Example 4:
• Consider the system:

– differentiate from 𝑦: 𝑦ሶ = 𝑥ሶ1 = 𝑥23 + 𝑢

– The control input is designed as: 𝑢 = −𝑥23 − 𝑒 𝑡 + 𝑦ሶ 𝑑 (𝑡)
– The internal dynamics is determined by 2nd equation:
𝑥ሶ 2 + 𝑥23 = 𝑦ሶ 𝑑 − 𝑒
– To keep the output equal zero: 𝑦𝑑 = 𝑦 = 𝑥1 = 0.
– The zero dynamics is
𝑥ሶ 2 + 𝑥23 = 0
– which 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
Feedback Linearization
• How to Generalize?

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
Contents
Introduction
1 Motivation examples on input-output, and input-state
feedback linearization, internal and zero dynamics.

Differential Geometry
2 Gradient and Jacobian of vector fields, Lie Derivative and Lie
bracket, involutivity, and diffeomorphism.

Input-Output Feedback Linearization

3 Relative degree, normal form, zero dynamics, non-minimum
phase systems, state feedback control, regulation and tracking.

Input-State Feedback Linearization

4 Feasibility conditions, suitable diffeomorphism, state-feedback
control, regulation and tracking.

In this chapter we review feedback linearization methods, as the most fundamental technique in the design of state feedback control for
nonlinear systems. The required mathematics and conditions are reviewed through a quick introduction to differential geometry. Then a
generalized methods for input-output and input-state feedback linearization.
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
Feedback Linearization
• Differential Geometry
 Consider the SISO nonlinear system
𝑥ሶ = 𝑓 𝑥 + 𝑔 𝑥 𝑢 (13.7)
𝑦=ℎ 𝑥 (13.8)
• where 𝑓, 𝑔: 𝑅𝑛 → 𝑅𝑛 are vector fields in 𝑅 𝑛
• and ℎ: 𝑅 𝑛 → 𝑅 is a scalar function.
 Gradient and Jacobian
• for ℎ: the gradient is defined as a row vector
𝜕ℎ 𝜕ℎ
– 𝛻ℎ = with elements: 𝛻ℎ 𝑗 =
𝜕𝑥 𝜕𝑥𝑗

• for 𝑓, 𝑔: the Jacobian is defined as an 𝑛 × 𝑛 Matrix:

𝜕𝑓 𝜕𝑓𝑖
– 𝛻𝑓 = with elements: 𝛻𝑓 𝑖𝑗 =
𝜕𝑥 𝜕𝑥𝑗

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
Differential Geometry
• Lie Derivative
 The time derivative 𝑦ሶ is given by:

𝜕ℎ
• Where 𝐿𝑓 ℎ 𝑥 = 𝑓(𝑥) is called
𝜕𝑥
The Lie Derivative of ℎ with respect to 𝑓
– The notion of derivative along the trajectories of system 𝑥ሶ = 𝑓 𝑥
• Repeated derivatives:

• Furthermore,
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
Differential Geometry
• Lie Bracket
 Let 𝑓, 𝑔 be two vector fields in 𝑅𝑛 . The Lie bracket of 𝑓 and 𝑔
is a third vector field defined by
𝑓, 𝑔 = 𝛻𝑔 𝑓 − 𝛻𝑓 𝑔
• Another notation:
𝜕𝑔 𝜕𝑓
𝑎𝑑𝑓 𝑔 = 𝑓− 𝑔
𝜕𝑥 𝜕𝑥
• Repeated adjoints:
𝑎𝑑𝑓0 𝑔 = 𝑔

𝑎𝑑𝑓𝑘 𝑔 = 𝑓, 𝑎𝑑𝑓𝑘−1 𝑔 , 𝑘≥1

• It is clear that 𝑓, 𝑔 = − 𝑔, 𝑓 .
• For constant vector fields 𝑓, 𝑔, 𝑓, 𝑔 = 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
 20
Differential Geometry
• Lie Bracket
 Example 1:
• Let 𝑓, 𝑔 be two vector fields in 𝑅2 as:

• Then:

• 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
 21
Differential Geometry
• Lie Bracket
 Example 2:
• Let 𝑓 𝑥 = 𝐴, and g is a constant vector field. Then

• And

 Properties:
• Bilinearity:
– If 𝑓, 𝑓1 , 𝑓2 , 𝑔, 𝑔1 , 𝑔2 are smooth vector fields, and 𝛼1 , 𝛼2 are constant scalars,
then

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
Differential Geometry
• Lie Bracket
 Properties:
• Jacobi Identity:
• Proof:
– LHS:

– RHS

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
Differential Geometry
• Involutivity
 For vector fields 𝑓1 , 𝑓2 , … , 𝑓𝑘 on 𝐷 ⊂ 𝑅𝑛 ,
• let ∆(𝑥) be the subspace of 𝑅𝑛 spanned by vectors:

• The dimension of ∆(𝑥) is defined as:

– This may vary with 𝑥

– But if 𝑓1 , 𝑓2 , … , 𝑓𝑘 are linearly independent for all 𝑥 ∈ 𝐷, then
dim ∆ 𝑥 = 𝑘
– In this case ∆ is a nonsingular distribution on 𝐷.
• The distribution ∆ is involutive if

• If ∆ is a nonsingular distribution on D, it is involutive if and only if

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
Differential Geometry
• Involutivity
 Notes:

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
Differential Geometry
• Involutivity
• Example 1:
– Let 𝐷 = 𝑅3 , and ∆= span{𝑓1 , 𝑓2 }, where

– It can be verified that dim ∆ = 2, ∀𝑥 ∈ 𝐷, and

– However,

– Hence ∆ is not involutive.

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
Differential Geometry
• Involutivity
• Example 2:
– Let

– It can be verified that dim ∆ = 2, ∀𝑥 ∈ 𝐷, and

– While

– Therefore 𝑓1 , 𝑓2 ∈ ∆, and ∆ is involutive.

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
Differential Geometry
• Diffeomorphism
 Definition:
• A continuously differentiable map 𝑧 = 𝑇(𝑥)
• With a continuously differentiable inverse 𝑥 = 𝑇 −1 (𝑧)
• Is called a diffeomorphism.

 Lemma:
• If the Jacobian matrix 𝛻𝑇 = 𝜕𝑇/𝜕𝑥 is nonsingular at a point
𝑥0 ∈ 𝐷, then by inverse function theorem, ∃ a neighborhood 𝑁
of 𝑥0 such that 𝑇 is a diffeomorphism on 𝑁.

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
Differential Geometry
• Diffeomorphism
 Example: Consider

– The Jacobian is:

– Has rank 2 @ 𝑥 = (0,0) 

– It is a local diffeomorphism.
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
Scientist Bio
Marius Sophus Lie
(17 December 1842 – 18 February 1899)

was a Norwegian mathematician. He largely created

the theory of continuous symmetry and applied it to
the study of geometry and differential equations. Lie
obtained his PhD at the University of Christiania
(present day Oslo) in 1871 with a thesis entitled On
a Class of Geometric Transformations. In 1886 Lie
became professor at Leipzig, he was made Honorary
Member of the London Mathematical Society in 1878,
Member of the French Academy of Sciences in 1892,
Foreign Member of the Royal Society of London in
1895 and foreign associate of the National Academy
of Sciences of the United States of America in 1895. Lie Groups
Lie died at the age of 56, due to pernicious anemia,
a disease caused by impaired absorption of vitamin
B12.
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
Contents
Introduction
1 Motivation examples on input-output, and input-state
feedback linearization, internal and zero dynamics.

Differential Geometry
2 Gradient and Jacobian of vector fields, Lie Derivative and Lie
bracket, involutivity, and diffeomorphism.

Input-Output Feedback Linearization

3 Relative degree, normal form, zero dynamics, non-minimum
phase systems, state feedback control, regulation and tracking.

Input-State Feedback Linearization

4 Feasibility conditions, suitable diffeomorphism, state-feedback
control, regulation and tracking.

In this chapter we review feedback linearization methods, as the most fundamental technique in the design of state feedback control for
nonlinear systems. The required mathematics and conditions are reviewed through a quick introduction to differential geometry. Then a
generalized methods for input-output and input-state feedback linearization.
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
Feedback Linearization
• Input-Output Feedback Linearization
 Reconsider the SISO nonlinear system
𝑥ሶ = 𝑓 𝑥 + 𝑔 𝑥 𝑢 13.7
𝑦=ℎ 𝑥 (13.8)
• Consider the time derivative 𝑦ሶ is given by:

• If 𝐿𝑔 ℎ 𝑥 = 0, then second derivative 𝑦 (2) is given

• Once, if 𝐿𝑔 𝐿𝑓 ℎ 𝑥 = 0, 𝑦 (2) = 𝐿𝑓2 ℎ(𝑥) is independent of 𝑢

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
Input-Output Feedback Linearization
• Repeat differentiation until u appears in 𝑦 (𝜌)

• Then u does not appear in 𝑦, 𝑦,ሶ … , 𝑦 (𝜌−1) , and appears in 𝑦 (𝜌)

• This results in the system to be input-output feedback linearizable,

and by the state feedback control

• The input-output map is reduced to:

• This is a chain of 𝜌 integrators.

• The integer 𝜌 is called the relative degree of the 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
 33
Input-Output Feedback Linearization

 Well-Defined Relative Degree

• Extension of Linear system excess poles or relative degree to
nonlinear system.
• Differentiate 𝜌 times 𝑦 until 𝑢 appears.

 Undefined Relative Degree

𝜌−1
• If 𝐿𝑔 𝐿𝑓 ℎ 𝑥 is zero at 𝑥0 , but nonzero at some points arbitrary
close to 𝑥0 ,
– Then the relative degree is undefined.
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
Input-Output Feedback Linearization
 Relative Degree
• Example 1: Consider the system 𝑥ሷ = 𝑔 𝑥, 𝑥ሶ + 𝑢
– With 𝑦 = 𝑥, the relative degree is 2, since:

𝑦ሶ = 𝑥,ሶ 𝑦ሷ = 𝑥ሷ = 𝑔 𝑥, 𝑥ሶ + 𝑢

– However, with 𝑦 = 𝑥 2 the relative degree is undefined:

𝑦ሶ = 2𝑥𝑥,ሶ 𝑦ሷ = 2𝑥𝑥ሷ + 2𝑥ሶ 2 = 2𝑥𝑔 + 2𝑥ሶ 2 + 2𝑥𝑢

Or 𝐿𝑔 𝐿𝑓 ℎ = 2𝑥

Therefore, @ 𝑥 = 0 this term is zero, but arbitrary close to 𝑥 = 0 this

term is nonzero  undefined relative degree.
– This can happen when 𝑢 ⋅ 𝑥 or 𝑢 ⋅ 𝑥ሶ appears in the derivatives of y.
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
Input-Output Feedback Linearization
 Relative Degree
• Example 2:
– Consider controlled Van der Pol

– With 𝑦 = 𝑥1 , the relative degree is 2.

– With 𝑦 = 𝑥2 , the relative degree is 1.

– With 𝑦 = 𝑥1 + 𝑥22 →
• The relative degree is 1 in
• The relative degree is undefined at 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
 36
Input-Output Feedback Linearization
 Relative Degree
• Example 3:
– Consider the system:

– Derivatives:

– Hence,

– The system has no well-defined relative degree.

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
Input-Output Feedback Linearization
 Relative Degree
• Example 4:
– Consider the field controlled DC motor:

– With 𝑦 = 𝑥3

– The relative degree is 2 in

– The relative degree is undefined at 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
 38
Input-Output Feedback Linearization
 Relative Degree and Excess Poles
• Example 5:
– Consider the linear system
With excess poles (𝑛 − 𝑚) > 0
– Controllable Cannonical Realization:

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
Input-Output Feedback Linearization
 Relative Degree and Excess Poles
• Example 5: (cont.)
– Derivatives:

– Otherwise, 𝐶𝐵 = 0 continue on differentiation:

In which,

– Hence, the relative degree is:

Which is the same as excess poles.

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
Input-Output Feedback Linearization
• Normal Form
 Consider a system with well defined relative degree 𝜌
• Use the new state variables, 𝜇 and 𝜓
• Through the diffeomorphism 𝑧 = 𝑇(𝑥) defined on 𝐷0
𝑧 = 𝑇 𝑥 = 𝜇1 𝜇2 … 𝜇𝜌 | 𝜓1 𝜓2 … 𝜓𝑛−𝜌 𝑇
In which,
𝑇 𝑇
𝜇 = 𝜇1 𝜇2 … 𝜇𝜌 = 𝑦 𝑦ሶ … 𝑦 (𝜌−1)
𝜌−1 𝑇
= ℎ 𝑥 𝐿𝑓 ℎ(𝑥) ⋯ 𝐿𝑓 ℎ(𝑥)
while the internal dynamics states 𝜓 shall satisfy
𝜕𝜓𝑖
𝑔 𝑥 = 0 for 1 ≤ 𝑖 ≤ 𝑛 − 𝜌, ∀𝑥 ∈ 𝐷0
𝜕𝑥

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
Input-Output Feedback Linearization
• Normal Form
 By use of these variables, the system is reduced into the
normal form:
𝜇2
⋮
𝜇ሶ = 𝜇𝜌
𝑎 𝜇, 𝜓 + 𝑏 𝜇, 𝜓 𝑢
𝜓ሶ = 𝑤(𝜇, 𝜓)
– With 𝑦 = 𝜇1 .
– 𝜇 and 𝜓 are called the normal states.
– 𝜇 is the external states, while 𝜓 represents the internal dynamics.
– Internal Dynamics is 𝜓ሶ = 𝑤(𝜇, 𝜓)
– Zero Dynamics is defined by: 𝜓ሶ = 𝑤(0, 𝜓)
– Feedback linearizable control is found by 𝑎 𝜇, 𝜓 + 𝑏 𝜇, 𝜓 𝑢 = 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
 42
Input-Output Feedback Linearization
• Normal Form
𝑇
 Note that 𝑧 = 𝜇 𝜓 = 𝑇(𝑥) shall be a diffeomorphism
• The 𝜇 state is canonical and linearly independent.
Lemma: If the relative degree of the system is 𝜌 in a domain 𝐷0 , then
the gradients 𝛻𝜇1 , 𝛻𝜇2 , … , 𝛻𝜇𝜌 are linearly dependent in 𝐷0 .

• The 𝜓 state shall complete the space

– The gradients 𝛻𝜓1 , 𝛻𝜓2 , … , 𝛻𝜓𝑛−𝜌 shall be linearly independent.
– And they must satisfy:
𝜕𝜓𝑖
𝛻𝜓𝑖 𝑔 𝑥 = 𝑔 𝑥 = 0 for 1 ≤ 𝑖 ≤ 𝑛 − 𝜌, ∀𝑥 ∈ 𝐷0
𝜕𝑥
This ensures that the input u does not influence the internal dynamics.
• Search for suitable 𝜓 that satisfies these conditions.

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
Input-Output Feedback Linearization
• Normal Form
 Example 1:
• Consider the system

– Derivatives:

– Hence, the relative degree is 2, and

– Find ψ to complete the space:

– One possible solution is:

Eq. point @
origin
– 𝑧 = 𝜇1 𝜇2 𝜓 is a diffeomorphism,
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
Input-Output Feedback Linearization
• Normal Form
• Example 1: (cont.)
– Since the Jacobian:
Is nonsingular for all x.
– Inverse Transformation:

– The Norm Form:

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
Input-Output Feedback Linearization
• Zero Dynamics
• Internal Dynamics in the normal form: 𝜓ሶ = 𝑤 𝜇, 𝜓
• For zero dynamics 𝑦 = 0 or 𝜇 = 0 in normal form.
• Hence, by definition the zero dynamics is:

𝜓ሶ = 𝑤 0, 𝜓

 Minimum Phase System

• If the origin of the zero dynamics is asymptotically stable in a
domain, the system is called minimum phase.
• If the system is minimum phase, I/O linearization keeps all the
states bounded.
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
Input-Output Feedback Linearization
 Zero Dynamics
• Example 1:
– Consider the system:

– The internal dynamics derived before as:

– The zero dynamics is (𝜇1 = 𝜇2 = 0)

– The system is minimum phase.

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
Input-Output Feedback Linearization
• Zero Dynamics
 Example 2: Controlled Van der Pol
• Consider the system as:

– Relative degree is 1.
– The system is originally in normal form, with
𝜇 = 𝑥2 , 𝜓 = 𝑥1 .
– Zero dynamics is:
𝜓ሶ = 𝑥ሶ1 = 0
– Zero dynamics does not have any equilibrium point
– The system is non-minimum phase!
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
Input-Output Feedback Linearization
• Zero Dynamics
 Example 3
• Consider the system:
– The eq. point is @ origin.

– Derivatives:
– The relative degree is 2.
– The normal coordinates are:
𝜇1 = 𝑥2 , 𝜇2 = 𝑥3 .
– The zero-dynamics is (𝑥2 = 𝑥3 = 0).
𝑥ሶ1 = 𝑥1
– The system is minimum phase.
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
Input-Output Feedback Linearization
• Zero Dynamics
 Example 3: (Cont.)
– To find 𝜓 and the internal dynamics, consider

and

– Hence,
– Solve by separation of variables:
– The normal form:

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
Input-Output Feedback Linearization
• Zero Dynamics
 Example 4:
• Consider the field controlled DC motor.
– The relative degree is 2, and 𝜇1 = 𝑥3 , 𝜇2 = 𝜃𝑥1 𝑥2 ,
𝜕𝜓
– 𝛻𝜓 ⋅ 𝑔 = 0 → ⋅ 1 = 0 → 𝜓 ≠ 𝑓(𝑥1 ) hence 𝜓 = 𝑓(𝑥2 )
𝜕𝑥1
– For zero dynamics consider: 𝜇1 = 𝜇2 = 0
– Hence, 𝑥ሶ 2 = −𝑏𝑥2 + 𝜌
– zero dynamics have an asymptotically stable equilibrium point at
𝑥2 = 𝜌/𝑏
– The system is minimum phase.
– choose 𝜓 = 𝑥2 − 𝜌/𝑏 to push the Eq. point @ 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
 51
Input-Output Feedback Linearization
• State Feedback Control 𝜇2
⋮
 Controller Design: 𝜇ሶ = 𝜇𝜌
• Reconsider the normal form 𝑎 𝜇, 𝜓 + 𝑏 𝜇, 𝜓 𝑢
– The 𝜇ሶ 𝑟 dynamic equation reads: 𝜓ሶ = 𝑤(𝜇, 𝜓)

𝜌 𝜌−1
𝑦 (𝜌) = 𝑎 𝜇, 𝜓 + 𝑏 𝜇, 𝜓 𝑢 = 𝐿𝑓 ℎ 𝑥 + 𝐿𝑔 𝐿𝑓 ℎ 𝑥 𝑢

– Design 𝑢 to linearize this equation to 𝑦 (𝜌) = 𝑣 by

1 1 𝜌
𝑢= −𝑎 + 𝑣 = 𝜌−1 (−𝐿𝑓 ℎ 𝑥 + 𝑣)
𝑏 𝐿𝑔 𝐿𝑓 ℎ 𝑥

• Stabilization
– Choose 𝑣 = −𝑘𝜌−1 𝑦 𝜌−1 − ⋯ − 𝑘1 𝑦ሶ − 𝑘0 𝑦
– such that the following polynomial is Hurwitz.
𝐾 𝑠 = 𝑠 𝜌 + 𝑘𝜌−1 𝑠 𝜌−1 + ⋯ + 𝑘1 𝑠 + 𝑘0 (13.9)
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
Input-Output Feedback Linearization
• State Feedback Control
 Stabilization
• The stabilizing control law is:
1 𝜌 𝜌−1
• 𝑢 𝑥 = 𝜌−1 [−𝐿𝑓 𝑦 − 𝑘𝜌−1 𝑦 − ⋯ − 𝑘1 𝑦ሶ − 𝑘0 𝑦]
𝐿𝑔 𝐿𝑓 𝑦

Theorem 13.1
Assume that the system (13.7),(13.8) has relative degree 𝜌, and
is minimum phase. Choose 𝑘𝑖 such that the polynomial (13.9) is
Hurwitz. Then the above mentioned control law yields to an
asymptotically stable closed-loop equilibrium point.

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
Input-Output Feedback Linearization
• State Feedback Control
 Tracking:
(𝜌−1) 𝑇
• Let us define: 𝜇𝑑 = 𝑦𝑑 𝑦ሶ 𝑑 … 𝑦𝑑
• And the tracking error as: 𝜇෤ 𝑡 = 𝜇 𝑡 − 𝜇𝑑 (𝑡)

Theorem 13.2
Assume that the system (13.7),(13.8) has relative degree 𝜌, that 𝜇𝑑 is smooth
and bounded, ant the solution 𝜓ሶ 𝑑 of the equation
𝜓ሶ 𝑑 = 𝑤 𝜇𝑑 , 𝜓𝑑 , 𝜓𝑑 0 = 0
exists, is bounded, and is uniformly asymptotically stable. Choose 𝑘𝑖 such that
the polynomial (13.9) is Hurwitz. Then the control law:
1 𝜌 (𝜌) 𝜌−1
𝑢 𝑥 = 𝜌−1 [−𝐿𝑓 𝑦 + 𝑦𝑑 − 𝑘𝜌−1 𝑦 − ⋯ − 𝑘1 𝑦ሶ − 𝑘0 𝑦]
𝐿𝑔 𝐿𝑓 𝜇1
The whole states remain bounded and the tracking error 𝜇෤ 𝑡 converges to
zero exponentially.

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
Input-Output Feedback Linearization
• State Feedback Control
 Example 1:
• Consider the system

– Choose the output as:

– Derivative:
– Zero dynamics:
𝑦 = 0 → 𝑥2 = − 2𝑥1 → 𝑥ሶ1 = −2𝑥13
– System in minimum phase.
– Stabilizing Control Law: 𝑢 = −2𝑥12 𝑥2 − 4𝑥2 − 2𝑥1
– Tracking Control Law: 𝑢 = −2𝑥12 𝑥2 − 3𝑥2 + 𝑦ሶ 𝑑 − 2(𝑦 − 𝑦𝑑 )
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
Input-Output Feedback Linearization
• State Feedback Control
 Example 2: (Pendulum)
• Control objective:
– Stabilize the pendulum @ 𝜃 = 𝛿1
• System dynamics

– In which,
– The output is chosen as 𝑦 = 𝑥1
y  x2 ; y  x2  a sin( x1  1 )  bx2  cu  v

– Stabilizing Control Law:

1
u a sin( x1  1 )  bx2  k1 x1  k2 x2 
c
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
Input-Output Feedback Linearization
• State Feedback Control
 Example 3:
• Consider the scalar system 𝑥ሶ = 𝑎𝑥 − 𝑏𝑥 3 + 𝑢
– A linearizing and stabilizing controller is
𝑢 = − 𝛾 + 𝑎 𝑥 + 𝑏𝑥 3 , 𝛾>0
– The closed loop system 𝑥ሶ = −𝛾𝑥
• This cancels the nonlinear damping −𝑏𝑥 3
– Another stabilizing controller:
• Do not cancel the damping
𝑢 = − 𝛾 + 𝑎 𝑥, 𝛾>0
3
– The closed loop system is 𝑥ሶ = −𝛾𝑥 − 𝑏𝑥
• Globally asymptotically stable.
• Simpler to implement.
• Faster in response.
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
Input-Output Feedback Linearization
• State Feedback Control
 Example 4: Field Controlled DC motor
• System dynamics

– The output is 𝑦 = 𝑥3 and 𝜇1 = 𝑥3 , 𝜇2 = 𝜃𝑥1 𝑥2 ,

– Derivatives:
𝑓(𝑥) +

– In which, f ( x)   (a  b) x2    cx1 x3 

– Stabilizing controller
1
u  f ( x)  k1 x3  k2 x1 x2  , x2  0
 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
 58
Input-Output Feedback Linearization
• State Feedback Control
 Example 5:
• Consider the system:
– The eq. point is @ origin.
– Derivatives:
– The relative degree is 2.
– The normal coordinates are:
𝜇1 = 𝑥2 , 𝜇2 = 𝑥3 .
– The stabilizing controller:
𝑢 = −𝑥1 𝑥3 − 𝑘1 𝑥2 − 𝑘2 𝑥3
– The tracking controller:
𝑢 = −𝑥1 𝑥3 + 𝑦ሷ 𝑑 − 𝑘1 (𝑥2 − 𝑦𝑑 ) − 𝑘2 (𝑥3 − 𝑦ሶ 𝑑 )
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
Contents
Introduction
1 Motivation examples on input-output, and input-state
feedback linearization, internal and zero dynamics.

Differential Geometry
2 Gradient and Jacobian of vector fields, Lie Derivative and Lie
bracket, involutivity, and diffeomorphism.

Input-Output Feedback Linearization

3 Relative degree, normal form, zero dynamics, non-minimum
phase systems, state feedback control, regulation and tracking.

Input-State Feedback Linearization

4 Feasibility conditions, suitable diffeomorphism, state-feedback
control, regulation and tracking.

In this chapter we review feedback linearization methods, as the most fundamental technique in the design of state feedback control for
nonlinear systems. The required mathematics and conditions are reviewed through a quick introduction to differential geometry. Then a
generalized methods for input-output and input-state feedback linearization.
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
Feedback Linearization
• Input-State Linearization
 Reconsider the single-input system
𝑥ሶ = 𝑓 𝑥 + 𝑔 𝑥 𝑢 13.7
Definition: A single input nonlinear system in the form of (13.7) with 𝑓, 𝑔 being
smooth vector fields on 𝑅𝑛 , is input-state feedback linearizable if there exists a
region 𝐷 in 𝑅𝑛 , a diffeomorphism 𝑧 = 𝑇(𝑥), 𝑇: 𝐷 → 𝑅𝑛 , and a nonlinear feedback
control law
𝑢 =𝛼 𝑥 +𝛽 𝑥 𝑣
Such that in the new state 𝑧, and the new input 𝑣 the system can be represented by
an LTI (controllable canonical) form of
𝑧ሶ = 𝐴𝑧 + 𝑏𝑣,
where:
0 1 0 ⋯ 0 0
0 0 1 0 0
𝐴= ⋮ ⋱ ⋱ ⋮ , 𝑏= ⋮ .
0 0 1 0
0 0 ⋯ 0 1
– Chain of n integrators.
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
Input-State Feedback Linearization
• Input-State Linearization
 Feasibility conditions:
• Input-output versus input-state condition

• Direct input-state conditions

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
Input-State Feedback Linearization
• Input-State Linearization
• Direct input-state conditions
• Note the first condition is like controllability:

𝐶 = [𝑏, 𝐴𝑏, …, 𝐴𝑛−1 𝑏]

• The second condition is just for nonlinear system, and assures

integrability of the state equations.
– Involutivity assures integrability

• How to generate the diffeomorphism and the control law.

– Proceed with the reciepe and some 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
 63
Input-State Feedback Linearization
• Input-State Linearization
 Find suitable diffeomorphism 𝑧 = 𝑇(𝑥)
• From the canonical form of the state 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
 64
Input-State Feedback Linearization
• Input-State Linearization
 Recipe

z1 zn 1
g ( x)  0, , g ( x)  0,
x x
zn
OR g ( x)  0
x

 z1 zn 1 
 z1 , z2  x f , , zn 
x 
f


 zn   z 
 ( x)    f ( x)  /  n g ( x) 
 x   x 
OR
 z 
 ( x)  1/  n g ( x) 
 x 
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
Input-State Feedback Linearization
• Input-State Linearization
 Example 1: Synchronous Generator
• System dynamics

– The open loop eq. point is @ origin.

– The diffeomorphism 𝑧 = 𝑇(𝑥) satisfies:

– With 𝑇1 (0) = 0 and

– From the equation

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
Input-State Feedback Linearization
• Input-State Linearization
 Example 1: (Cont.)
– Choose 𝑇1 independent of 𝑥3 ; Therefore,

– From the condition:

– Choose 𝑇1 independent of 𝑥2 ; Therefore,

– and,

– Therefore,

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
Input-State Feedback Linearization
• Input-State Linearization
 Example 1: (Cont.)
– Choose 𝑇1 = 𝑥1 , Hence:

– And the inverse transformation is:

– The controller is: 𝑢 = 𝛼 𝑥 + 𝛽 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
 68
Input-State Feedback Linearization
• Input-State Linearization
 Example 1: (Cont.)
• The normal form:

• The linearizing controller:

• The feedback linearizing is locally valid in the domain:

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
Input-State Feedback Linearization
• Input-State Linearization
 Example 2: Flexible Link Manipulator
• System Dynamics:
– With

– Where 𝑎, 𝑏, 𝑐, 𝑑 are constant parameters.

• Feedback Linearizable?
– Involutivity: The vector field
therefore, involutive.

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
Input-State Feedback Linearization
• Input-State Linearization
 Example 2: (Cont.)
• Feedback Linearizable?
– Controllability: The rank of:

– Is four ∀𝑥, therefore, the system is controllable.

• Diffeomorphism:
– Find ℎ that satisfies:

𝜕ℎ 𝜕ℎ
– From the condition 𝑔 = 0 we have = 0 so ℎ is independent of 𝑥4 .
𝜕𝑥 𝜕𝑥4
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
Input-State Feedback Linearization
• Input-State Linearization
 Example 2: (Cont.)
– Thus,
𝜕(𝐿𝑓 ℎ) 𝜕(𝐿𝑓 ℎ) 𝜕ℎ
– From the condition 𝑔 = 0 we have = 0, so = 0 and
𝜕𝑥 𝜕𝑥4 𝜕𝑥3
ℎ is independent of 𝑥3 .
– This simplifies:
– Now move forward:

– Therefore,

– Choose ℎ is independent of 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
 72
Input-State Feedback Linearization
• Input-State Linearization
 Example 2: (Cont.)
– Hence:

– To assure , ℎ shall be a function of 𝑥1 .

– Choose the simplest function ℎ 𝑥 = 𝑥1 . Then

– The final state coordinate is globally defined as:

– 𝑧1ሶ = 𝑧2 , 𝑧2ሶ = 𝑧3 , 𝑧3ሶ = 𝑧4 , 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
 73
Input-State Feedback Linearization
• Input-State Linearization
 Example 2: (Cont.)
• The inverse of the transformation is globally defined by:
– 𝑥1 = 𝑧1
– 𝑥2 = 𝑧2
1
– 𝑥3 = 𝑧1 + 𝑧3 + 𝑎 sin 𝑧1
𝑏
1
– 𝑥4 = 𝑧2 + 𝑧4 + 𝑎𝑧2 cos 𝑧1
𝑏

• The controller is: 𝑢 = 𝛼 𝑥 + 𝛽 𝑥 𝑣 = (𝑣 − 𝛻𝑧4 𝑓)/(𝛻𝑧4 𝑔)

1
• Thus, 𝑢 = [𝑣 −𝛾 𝑥 ]
𝑏𝑑
– where,
𝛾 𝑥 = 𝑎𝑥22 sin 𝑥1 + acos 𝑥1 + 𝑏 asin 𝑥1 + 𝑏 𝑥1 − 𝑥3 + 𝑏𝑐 𝑥1 − 𝑥3 .
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
Input-State Feedback Linearization
• Input-State Linearization
 Example 3: Field-controlled DC motor
• System dynamics

– Where 𝑎, 𝑏, 𝑐, 𝜃, 𝑘 are positive constants.

– Check feedback linearization conditions:

– The determinant of:

– Is given by
– Hence, it has rank three for 𝑥2 ≠ 𝑘/2𝑏 and 𝑥3 ≠ 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
 75
Input-State Feedback Linearization
• Input-State Linearization
 Example 3: (Cont.)
• Involutivity
– It is involutive.
– The system is input-state feedback linearizable in the domain:

• Find suitable diffeomorphism 𝑧 = ℎ(𝑥)

– The unforced equilibrium point is 𝑥1∗ = 0, 𝑥2∗ = 𝑘/𝑏.
– Let 𝑥3∗ = 𝜔0 , where 𝜔0 is the desired set point for 𝑥3 .
– Diffeomorphism first 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
 76
Input-State Feedback Linearization
• Input-State Linearization
 Example 3: (Cont.)

– From ℎ must be independent of 𝑥1 , therefore

– From the condition we have

– This is satisfied if ℎ is taken as

– For some constants 𝑐1 , 𝑐2 . Choose 𝑐1 = 1, and satisfy ℎ 𝑥 ∗ = 0

– Then

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
Input-State Feedback Linearization
• Input-State Linearization
 Example 3: (Cont.)
– Hence:

– The condition

– Is satisfied, if 𝑥2 ≠ 𝑘/2𝑏, and 𝑥3 ≠ 0.

– Assume 𝑥3∗ > 0, then 𝑧 = 𝑇(𝑥) is a diffeomorphism on 𝐷0

– Where 𝜙(⋅) is the inverse map of 2𝜃𝑥2 𝑘 − 𝑏𝑥2 .

– Which is well defined for 𝑥2 > 𝑘/2𝑏. 𝐷𝑧 contains 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
 78
Input-State Feedback Linearization
• Input-State Linearization
 Example 3: (Cont.)
• Choose:
– 𝑇1 = ℎ = 𝜃𝑥22 + 𝑐𝑥32 + 𝑐2 , Hence:
𝜕𝑇1
– 𝑇2 = 𝑓 = 2𝜃𝑥2 [−𝑏𝑥2 + 𝑘],
𝜕𝑥
𝜕𝑇2
– 𝑇3 = 𝑓 = 2𝜃 −2𝑏𝑥2 + 𝑘 −𝑏𝑥2 + 𝑘 − 𝑐𝑥1 𝑥3 .
𝜕𝑥

• The controller is: 𝑢 = 𝛼 𝑥 + 𝛽 𝑥 𝑣

– where,
1 1 1
– 𝛽 𝑥 = 𝜕𝑇3 = 𝜕𝑇3 =
𝑔 2𝜃𝑐𝑥3 [2𝑏𝑥2 −𝑘]
𝜕𝑥 𝜕𝑥1
𝜕𝑇3 𝜕𝑇3
𝑓 𝑓
– 𝛼 𝑥 =− 𝜕𝑥
𝜕𝑇3 = 𝜕𝑥
=⋯
𝑔 2𝜃𝑐𝑥3 2𝑏𝑥2 −𝑘
𝜕𝑥

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 6: Feedback Linearization
To read more and see the course videos
visit our course website:
[Link]

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 [Link]. degree in mechanical engineering
from Sharif University of Technology, Tehran, Iran, in 1989, his [Link]. 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