ECE 2680: Adaptive Control (3 Credits, Fall 2018)

Lecture 1: Course Organization and

Introduction to Adaptive Control

August 30, 2018

Zhi-Hong Mao
Professor of ECE and Bioengineering
University of Pittsburgh, Pittsburgh, PA

Outline

• Course description
• Course organization
• Why adaptive control?
• What is adaptive control?
• Approaches to adaptive control

Course description
• Why I decided to develop this course?
• This course introduces the general principles of
adaptive control and learning

ECE 1673: Linear control systems

ECE 2646: Linear system theory

Nonlinear control Optimal control

Robust control Adaptive control …

1
 Course description
• Why I decided to develop this course?
• This course introduces the general principles of adaptive control and learning

• This course covers:

– System identification and real-time parameter estimation
– Model-reference adaptive systems
– Self-tuning regulators
– Introduction to computational learning theory and learning
in neural systems

Course organization

• Time: Thursday 5:20 pm-7:50 pm

• Instructor: Dr. Zhi-Hong Mao
– (Office) 1204 Benedum Hall
– (Email) [email protected]
– (Phone) 412-624-9674
– (Office hours) Monday 3:30 pm−5:30 pm

Course organization
• Time
• Instructor

• Text book
– S. Sastry and M. Bodson, Adaptive Control: Stability,
Convergence, and Robustness, Prentice-Hall, 1989-1994,
Sastry & Bodson, 1994, available for free download at
http://www.ece.utah.edu/~bodson/acscr/
• Lecture notes
• Email list
– Important or emergent notice will be sent to you via emails

2
 Course organization
• Time
• Instructor
• Text book
• Lecture notes
• Email list

• Course evaluation
– Homework: 25% (late homework will not be accepted)
– Class participation: 10%
– Midterm: 25%
– Final exam: 40%

Why adaptive control?

• Systems to be controlled have parameter

uncertainty
• System dynamics experience unpredictable
parameter variations as the control operation
goes on

Why adaptive control?

• Systems to be controlled have parameter

uncertainty
• System dynamics experience unpredictable
parameter variations as the control operation
goes on

• Examples
– Robot manipulation

3
 Why adaptive control?

• Systems to be controlled have parameter

uncertainty
• System dynamics experience unpredictable
parameter variations as the control operation
goes on

• Examples
– Robot manipulation

– Ship steering

10

Why adaptive control?

• Systems to be controlled have parameter uncertainty
• System dynamics experience unpredictable parameter variations as the
control operation goes on

• Examples
– Robot manipulation
– Ship steering

– Aircraft control

11

Why adaptive control?

• Systems to be controlled have
parameter uncertainty
• System dynamics experience
unpredictable parameter variations
as the control operation goes on

• Examples
– Robot manipulation
– Ship steering
– Aircraft control

– Process control

12

4
 Why adaptive control?
• Systems to be controlled have
parameter uncertainty
• System dynamics experience
unpredictable parameter variations
as the control operation goes on

• Examples
– Robot manipulation
– Ship steering
– Aircraft control
– Process control

– Neural control of
movement

13

What is adaptive control?

• Definition
– A form of control in which the control parameters are
automatically adjusted as conditions change so as to
optimize performance [cited from Oxford English
Dictionary]

14

What is adaptive control?

• Definition

• Difference between adaptive control and

ordinary feedback control

Reference
signal Controller Plant Output
Control
signal

15

5
 What is adaptive control?

• Definition

• Difference between adaptive control and

ordinary feedback control
Parameter
adjustment
Controller
parameters
Reference
signal Controller Plant Output
Control
signal

16

What is adaptive control?

• Definition
• Difference between adaptive control and ordinary feedback control

• Difference between adaptive control and robust

control
– Robust control can also be used to deal with parameter
uncertainty
– Adaptive control is superior to robust control in dealing
with uncertainties in constant or slowly-varying parameters
– Robust control has some desirable features which
adaptive control does not have, such as its ability to deal
with quickly varying parameters and unmodeled dynamics

17

What is adaptive control?

• Definition
• Difference between adaptive control and ordinary feedback control
• Difference between adaptive control and robust control

• A brief history
– In 1950s, connection with design of autopilots for
high-performance aircraft
“My interest, of course, stems from the very fundamental relationships
that exist in all our military weapon systems. I like to use the analogy of the
three-legged milking stool, with the seat representing the warhead: one leg
representing aerodynamics; another leg, propulsion systems; and the third
leg representing guidance and control. Without any of these legs you don’t
have an effective military weapon….
It is my feeling that we don’t have a corresponding amount of effort on
this other leg of the stool representing guidance and control including the
adaptive control systems.” –Maj. Gen. L. I. Davis, 1959 18

6
 What is adaptive control?
• Definition
• Difference between adaptive control and ordinary feedback control
• Difference between adaptive control and robust control

• A brief history
– In 1950s, connection with design of autopilots for
high-performance aircraft

F101A (VooDoo)

19

What is adaptive control?

• Definition
• Difference between adaptive control and ordinary feedback control
• Difference between adaptive control and robust control

• A brief history
– In 1950s, connection with design of autopilots for
high-performance aircraft
• During that time, the work on adaptive control was
characterized by a “lot of enthusiasm, bad hardware
and nonexisting theory” [Astrom, 1983]. The interest in
adaptive control soon diminished due to the lack of
insights and stability proofs coupled by the crash of a
test flight

20

What is adaptive control?

• Definition
• Difference between adaptive control and ordinary feedback control
• Difference between adaptive control and robust control

• A brief history
– In 1950s, connection with design of autopilots for high-performance aircraft

– In late 1960s, advances in control theory contributing

to the development of adaptive control (e.g., state-
space description, Lyapunov’s stability theory, and
dynamic programming)

21

7
 What is adaptive control?

• Definition
• Difference between adaptive control and ordinary feedback control
• Difference between adaptive control and robust control

• A brief history
– In 1950s, connection with design of autopilots for high-performance aircraft
– In 1960s, advances in control theory contributing to the development of adaptive control
(e.g., state-space description, stability theory, and dynamic programming)

– In 1970s and early 1980s, proofs for stability of

adaptive systems
– In late 1980s and early 1990s, new insights into
robustness of adaptive controllers
– Recently, strong relations to ideas on learning

22

Approaches to adaptive control

• Gain scheduling

23

Approaches to adaptive control

• Gain scheduling
– Advantages of gain scheduling
• Parameters can be changed quickly in response to changes
in the plant dynamics. It is convenient especially if the plant
dynamics depend in a well-known fashion on a relatively few
easily measurable variables

24

8
 Approaches to adaptive control

• Gain scheduling
– Advantages of gain scheduling

– Disadvantages of gain scheduling

• It is an open-loop adaptation scheme, with no real “learning”
or intelligence
• The extent of design required for its implementation can be
enormous (example: flight control system of CH-47
helicopter)

25

Approaches to adaptive control

• Gain scheduling

CH-47
(Chinook)

26

Approaches to adaptive control

• Gain scheduling

Multiple paired
internal models
in the
cerebellum

27

9
 Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme

28

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme

Pˆ ( s )

Questions: What is the transfer function between ym

and yp? Why use high gain k? 29

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme

Pˆ ( s )
k
_

What is the transfer

Pˆ ( s ) function of this block?

Using high-gain and a plant model to implement

an inverse model of the plant
30

10
 Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme
• Advantages: intuitive and simple—only one parameter
is updated

31

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme
• Advantages

• Disadvantages:
(1) Oscillations are constantly present in the system
(2) Noise in the frequency band of the limit cycle detector
causes the gain to decrease well below the critical value
(3) Reference inputs may cause saturation due to the high-gain
(4) Saturation may mask limit cycle oscillations, allowing the
gain to increase above the critical value, and leading to
instability

32

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme

33

11
 Approaches to adaptive control
• Gain scheduling

• Model reference adaptive systems

– Series high-gain scheme

– Parallel scheme

34

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive

systems
– Series high-gain scheme

– Parallel scheme
• Two loops: (1) an inner or
regulator loop that is an
ordinary control loop consisting
of the plant and regulator, and
(2) an outer or adaptation loop
that adjusts the parameters of
the regulator in such a way as
to drive the error between the
model output and plant output
to zero 35

Approaches to adaptive control

• Gain scheduling

• Model reference adaptive

systems
– Series high-gain scheme

– Parallel scheme
• Too loops

• Adjustment mechanism: driving

the output error e0 = yp− ym to
zero
Gradient update
d  2 Sensitivity function
dt
 g

 e0 ( )
 
 2 ge0 ( )  e0 ( )   2 ge0 ( )  y p ( ) 
 
Adaptation gain 36

12
 Approaches to adaptive control
• Gain scheduling

• Model reference adaptive

systems
– Series high-gain scheme

– Parallel scheme
• Too loops

• Adjustment mechanism: driving

the output error e0 = yp− ym to
zero
d 
Gradient update  2 ge0 ( )  y p ( ) 
dt 

M.I.T. rule: an implementation of the gradient update by

replacing the unknown parameters in ∂yp()/∂ by their
estimates at time t
37

Approaches to adaptive control

• Gain scheduling
• Model reference adaptive systems

• Self tuning regulators

– Certainty equivalence principle

38

Approaches to adaptive control

• Gain scheduling
• Model reference adaptive systems

• Self tuning regulators

– Certainty equivalence principle

– Two loops
• Inner loop consisting of a
conventional controller,
but with varying
parameters
• Outer loop consisting of
an identifier and design
box which adjust these
controller parameters
39

13
 Approaches to adaptive control
• Gain scheduling
• Model reference adaptive systems

• Self tuning regulators

– Certainty equivalence principle
– Two loops

– Direct and indirect adaptive control

40

Approaches to adaptive control

• Gain scheduling
• Model reference adaptive systems
• Self tuning regulators

• Stochastic control approach

41

References
• K. J. Astrom, Theory and applications of adaptive control—a survey, Automatica, vol. 19,
no. 5, pp. 471-486, 1983.
• K. J. Astrom, Adaptive control around 1960, IEEE Control Systems, pp. 44-49, June 1996.
• K. J. Astrom and B. Wittenmark, Adaptive Control, 2nd Edition, Addison-Wesley, 1995.
• R. C. Miall, D. J. Weir, D. M. Wolpert, and J. F. Stein, Is the cerebellum a Smith predictor?
Journal of Motor Behavior, vol. 25, no. 3, pp. 303-216, 1993.
• S. Sastry and M. Bodson, Adaptive Control: Stability, Convergence, and Robustness,
Prentice-Hall, 1989.
• J.-J. E. Slotine and W. Li, Applied Nonlinear Control, Prentice Hall, 1991.
• D. M. Wolpert, R. C. Miall, and M. Kawato, Internal models in the cerebellum, Trends in
Cognitive Sciences, vol. 2, no. 9, pp. 338-347, 1998.
• http://www.canlines.com/images/control_int/bluescreen.jpg
• http://www.geocities.com/Hollywood/2307/ch-47d_chinook.jpg
• http://www.infovisual.info/03/038_en.html
• http://www.photosfan.com/images/stealth-aircraft-flying-in-formation1.jpg
• http://www.robotokyo.com/wp-content/uploads/2007/04/domo_robot.jpg
• http://www.shareholdervalue.com/images/cartoon_art/Steering_Ship.jpg
• http://www.surclaro.com/photo6937.html

42

14