A course on:

Adaptive Control
Methods, algorithms and applications
From needs to applications
I.D. Landau, E. Witrant, GIPSA-Lab,Dept. of Automatic Control
CNRS / UGA

MiSCIT, 2021

1
I.D. Landau, : « A course on Adaptive Control »- 1
 Outline of the Course

1. Introduction
2. Parameter adaptation algorithms
3. Identification in open loop – a brief review
4. Iterative identification in closed loop and controller redesign
5. Direct and Indirect Adaptive Control
6. Robust control design for Adaptive Control
7. Parameter estimators for Adaptive Control
8. Adaptive control with multiple models
9. Adaptive regulation (versus adaptive control)
10.Adaptive feedforward compensation of disturbances

2
I.D. Landau, : « A course on Adaptive Control »- 1
Adaptive Control

Part 1: Introduction

3
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control

A set of techniques for automatic adjustment of the controllers

in real time, in order to achieve or to maintain a desired level
of performance of the control system when the parameters of the
plant (disturbance) dynamic model are unknown and/or change
in time

Particular cases:
1 Automatic tuning of the controllers for unknown but constant plant parameters
2 Unpredictable change of the plant (disturbance) model in time

4
I.D. Landau, : « A course on Adaptive Control »- 1
 Outline
• Concepts
• Basic schemes
• Adaptive control versus Robust control
• Adaptive control configurations
(open loop adaptation, direct and indirect adaptive control)
• Parameter adaptation algorithms
• RST digital controller
• Adaptive control: regimes of operation
• Identification in closed and controller redesign
• Use of a priori available information
• Adaptive control with multiple models
• Adaptive regulation
• Adaptive feedforward compensation of disturbances
•Example of applications
5
I.D. Landau, : « A course on Adaptive Control »- 1
 Conceptual Structures
Desired
Performance Controller Plant
Design Model
Method Principle of model based
Reference
control design
u y
Controller Plant

Desired
Performance Adaptation
Scheme

An adaptive control
Reference structure
u y
Controller Plant

Remark:
An adaptive control system is nonlinear since controller parameters will depend upon u and y
6
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive control- Why ?

• High performance control systems may require precise tuning

of the controller but plant (disturbance) model parameters may
be unknown or time-varying
• “Adaptive Control” techniques provide a systematic approach for
automatic on-line tuning of controller parameters
• “Adaptive Control” techniques can be viewed as approximations
of some nonlinear stochastic control problems (not solvable in
practice)
• Objective of “Adaptive Control” : to achieve and to maintain
acceptable level of performance when plant (disturbance) model
parameters are unknown or vary

7
I.D. Landau, : « A course on Adaptive Control »- 1
Adaptive Control versus Conventional Feedback Control

Disturbances

Acting upon Acting upon the plant

controlled variables model parameters
(modifying control system perf.)

How to reduce the

effect of disturbances ?
Conventional Adaptive
feedback control control

Measurement: Measurement:
Controlled variables Index of performance (I.P.)

8
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control – Basic Configuration
Adjustable control system

Distrubances

Adjustable
Controller Plant

Desired
Performance

Comparison-
Decision Adaptation Performance
Mechanism Measurement

Adaptation scheme

9
I.D. Landau, : « A course on Adaptive Control »- 1
Adaptive Control versus Conventional Feedback Control

Conventional Feedback Adaptive Control

Control System System
Obj.: Monitoring of the “controlled” Obj.: Monitoring of the
variables according to a certain index performance (IP) of the control
of performance (IP) for the case of system for unknown and varying
known parameters parameters

Meas.: Controlled variables Meas.: Index of performance (IP)

Transducer IP measurement

Reference output Desired IP

Comparison block Comparison decision block

Controller Adaptation mechanism

10
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control versus Conventional Feedback Control

A conventional feedback control system is mainly dedicated to

the elimination of the effect of disturbances upon the controlled
variables.

An adaptive control system is mainly dedicated to the elimination

of the effect of parameter disturbances (variations) upon the
performance of the control system.

Adaptive control system = hierarchical system:

• Level 1 : Conventional feedback system
• Level 2 : Adaptation loop

11
I.D. Landau, : « A course on Adaptive Control »- 1
 Fundamental Hypothesis in Adaptive Control

For any possible values of plant (disturbance) model parameters

there is a controller with a fixed structure and complexity such
that the specified performances can be achieved with appropriate
values of the controller parameters

The task of the adaptation loop is solely to search for the “good”
values of the controller parameters

12
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control versus Robust Control

Adaptive control can further improve the performance of a

robust control system by:
• expanding the range of uncertainty for which performance
specification can be achieved
• better tuning of the nominal controller

For building an adaptive control systems robustness issues for

the underlying controller design can not be ignored.

The objective is to add adaptation capabilities to a robust controller

and not to use adaptive approach for tuning a non robust controller.

13
I.D. Landau, : « A course on Adaptive Control »- 1
Conventional Control – Adaptive Control - Robust Control
Parameters change Parameters change

Conventional versus Adaptive Conventional versus Robust

14
I.D. Landau, : « A course on Adaptive Control »- 1
Conventional Control – Adaptive Control - Robust Control

Conventional LC
+
Adaptation

Conventional LC
+
Robust Adaptation

Robust LC
+
Robust Adaptation

LC = linear controller

Robust Adaptive Control and Adaptive Robust Control

are different.
What we need : Robust Adaptation of a Robust Controller
15
I.D. Landau, : « A course on Adaptive Control »- 1
Basic Adaptive Control Configurations

16
I.D. Landau, : « A course on Adaptive Control »- 1
 Digital Control System

The control law is implemented on a digital computer

e(k) u(k) u(t) PLANT y(t) y(k)

r(k)
DAC
+ DIGITAL
COMPUTER + Actuator Process Sensor ADC
- ZOH

CLOCK

ADC: analog to digital converter

DAC: digital to analog converter
ZOH: zero order hold

17
I.D. Landau, : « A course on Adaptive Control »- 1
 Digital Control System

CLOCK

e(k) u(k) u(t) y(t) y(k)

r(k) DAC
+
COMPUTER + PLANT ADC
- ZOH

DISCRETIZED PLANT

- Sampling time depends on the

plant bandwidth !
- Efficient use of computer resources

The adaptive control algorithms will be directly derived in discrete-time form.

They are not discretizations of continuous time adaptive control solutions

18
I.D. Landau, : « A course on Adaptive Control »- 1
 The R-S-T Digital Controller

CLOCK

r(t) u(t)
D/A y(t)
Computer + PLANT A/D
(controller) ZOH

Discretized Plant

r(t) Bm u(t)
B y(t)
−d
T (q )
+ 1 q
Am S ( q) A
-
Plant
R (q ) Model
Controller
−1
q y (t )=y (t−1)
19
I.D. Landau, : « A course on Adaptive Control »- 1
 Open Loop Adaptive Control

ENVIRONMENT

ADJUSTABLE PLANT
CONTROLLER
+
-

ENVIRONMENT
TABLE
MEASURE

Assumption: known and rigid relationship between some

measurable variables (characterizing the environment) and
the plant model parameters
Called also: gain-scheduling systems
20
I.D. Landau, : « A course on Adaptive Control »- 1
 Indirect Adaptive Control
SUPERVISION ADAPTATION
LOOP

PERFORMANCE CONTROLLER PLANT

COMPUTATION MODEL
SPECIFICATIONS
ESTIMATION

ADJUSTABLE u y
CONTROLLER PLANT
+
-
Prediction (adaptation)
error
u(t) y(t)
Plant
Plant Model
+
Parametric
ε
Basic Estimation q -1 Adaptation
Scheme - Algorithm

Adjustable
^y (t )
Predictor

21
I.D. Landau, : « A course on Adaptive Control »- 1
 Direct Adaptive Control
(model reference adaptive control)
The reference model gives the desired time trajectory of the plant output
ADAPTATION LOOP

REFERENCE
MODEL - + Adaptation
ε
error
PARAMETRIC
ADAPTATION
ALGORITHM

ADJUSTABLE u y
+
PLANT
-
CONTROLLER

Resemblance with plant parameter estimation scheme

Reference model Plant

Adjustable feedback syst. Adjustable predictor
22
I.D. Landau, : « A course on Adaptive Control »- 1
 How to get a Direct Adaptive Control scheme ?
• Express the performance error in term of difference between
the parameters of an unknown optimal controller and those
of the adjustable controller

• Re-parametrize indirect adaptive control scheme (if possible)

such that the adaptive predictor will provide directly the
estimated parameters of the controller.
See : Adaptive Control (Landau, Lozano, M’Saad, Karimi) pg. 16

The number of situations for which a direct adaptive control

scheme can be developed is limited.

23
I.D. Landau, : « A course on Adaptive Control »- 1
 Identification in Closed Loop and Adaptive Control

• Identification in closed loop operation using appropriate

algorithms provides better models for design
• An iterative approach combining identification in closed loop
followed by a re-design of the controller is a very powerful
(auto-)tuning scheme
noise
r u + y
Controller Plant + +
- CLOE Algorithms
û ^y
Controller Model Closed Loop Output Error

Simulated System
Parametric
Adaptation
Algorithm
24
I.D. Landau, : « A course on Adaptive Control »- 1
 Iterative Identification in Closed Loop
and Controller Re-Design
w
r
T
+
1/S
u q-d B/A + y
- +
Plant + CL
R
+ u y -
1/S q-d B/A

- Model
R

Step 1 : Identification in Closed Loop

-Keep controller constant
-Identify a new model such that CL

Step 2 : Controller Re – Design

- Compute a new controller such that CL
Repeat 1, 2, 1, 2, 1, 2,…
25
I.D. Landau, : « A course on Adaptive Control »- 1
 Iterative Identification and Controller Redesign versus
(Indirect) Adaptive Control

Parameter Estimation
+ N = 1 : Adaptive Control
Controller Computation
t t+1 time
N = Small
Adaptive Control
Fixed (or time varying) N = Large
Controller
Controller computed at( t)
computed
Iterative Identification in C.L.
+
Parameter Estimation at (t +N) And Controller Re-design
N ⇒∞
t t+N time
Plant Identification in C.L. +
Controller Re-design
The iterative procedure introduces a time scale separation
between identification and control
26
I.D. Landau, : « A course on Adaptive Control »- 1
 Parametric adaptation algorithm (PAA)

Parameter vector θ = contains all the parameters of the model (or of the controller)

New parameters Old parameters

[ estimation
( vector ) ][
= estimation
(vector )
+
]
Adaptation Measurement Error prediction

[ Gain
( matrix ) ][
× function
(vector )
× function
(scalar ) ][ ]
Regressor vector
Estimated
Parameter ^
vector
θ(t+1 )=θ^ (t )+FΦ(t )v(t+1) (v=f ( ε))

27
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control Schemes. Regimes of operation

• Adaptive regime
1. Controller parameters are updated at every sampling time
2. Plant parameters are estimated at every sampling time but
controller parameters are updated only every N samples (N small)
3 Adaptation works only when there is enough excitation

• Self-tuning regime (parameters are supposed unknown but constant)

1 Parameter adaptation algorithms with decreasing adaptation gain
2 Controller parameters are either updated at every sampling time
or kept constant during parameter estimation
3 An external excitation is applied during tuning or plant identification
Remark
If controller parameters are kept constant during parameter estimation this is called
“auto-tuning”. For the indirect approach this corresponds to “plant identification
in closed loop operation and controller redesign”
28
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control and Adaptive Regulation

Adaptive Control
Plant model is unknown and time varying
The disturbance model is known and constant

Adaptive Regulation
Plant model is known and constant
The disturbance model is unknown and time varying

Adaptive control and regulation

Very difficult problem since is extremely hard to distinguish in the
performance (prediction) error what comes from plant model
error and what comes from disturbance model error

Remark:
The “internal model principle” has to be used in all the cases
29
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control δ (t ) : Dirac
δ (t )(or e(t )) e(t ) : White noise
disturbance source (unmeasurable)

Disturbance unmeasurable
input output model disturbance
PLANT
u(t ) Plant
+ p(t ) y(t )
model +

Objective : tracking/disturbance attenuation performance

• Focus on adaptation with respect to plant model parameters variations

• The model of the disturbance is assumed to be known and constant
• Only a level of attenuation in a frequency band is required*
• No effort is made to simultaneously estimate the model of the disturbance

*) Except for known DC disturbances (use of integrators)

30
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Regulation
δ (t )(or e(t ))
disturbance source (unmeasurable)

Disturbance unmeasurable
input output model disturbance
PLANT
u(t ) Plant
+ p(t ) y(t )
model +

Objective : Suppressing the effect of the (unknown) disturbance*

• Focus on adaptation with respect to disturbance model parameters
variations
• Plant model is assumed to be known ( a priori system identification)
and almost constant
• Small plant parameters variations handled by a robust control design
• No effort is made to simultaneously estimate the plant model

*) Assumed to be characterized by a rational power spectrum if stationary 31

I.D. Landau, : « A course on Adaptive Control »- 1
Use of a priori information for improving adaptation transients

- Before using an adaptive control scheme, an analysis of the system

is done and this is followed by plant identification in various
regimes of operation
- The availability of models for various regimes of operation allows
to design robust controllers which can assure satisfactory
performance in a region of the parameter space around each of
the identified models.
- Provided that we can detect in what region the system is, the
appropriate controller can be used
- “Indirect adaptive control” can not detect quickly enough the region
of operation but can make a “fine” tuning over a certain time.
- In case of rapid parameter changes the adaptation transients in
indirect adaptive control may be unacceptable.
- There is a need to improve these transients by taking in account
the available information 32
I.D. Landau, : « A course on Adaptive Control »- 1
 Supervisory Control
MODELS
- 
G1
+

-
G2
CONTROLLERS .. +
n
K1 -
Gn
SUPERVISOR
+

K2 PLANT
.. y

Kn

The “supervisor”:
• will check what “plant-model” error is minimum
• will switch to the controller associated with the selected model
Can provide a very fast decision (if there are not too many models)
but not a fine tuning
33
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control with Multiple Models

-
G
Adaptive model +

-
G1
Fixed models +

-
G2
.. +
-
n
Gn
SUPERVISOR
r +

Controller PLANT
u y
+ CL

-
Controller G y
u
P.A.A.

The supervisor select the best fixed model and then the adaptive model will be selected
Multiple fixed models : improvement of the adaptation transients
Adaptive plant model estimator (CLOE Estimator) : performance improvement
34
I.D. Landau, : « A course on Adaptive Control »- 1
Some Applications of Adaptive Control

35
I.D. Landau, : « A course on Adaptive Control »- 1
 Open Loop Adaptive Control of Deposited Zinc in
Hot-Dip Galvanizing

Finished
Steel product
strip air air
oven

Air Measurement of

Preheat knives deposited mass

oven
zinc
Steel strip

Zinc bath

input: air knives pressure Ge

−sτ
L L- distance
output: measured deposited mass H ( s)= ; τ=
1+sT V knives –measure
V- strip speed
• delay varies with the speed
• G and T depend upon strip speed and distance between knives and steel strip
36
I.D. Landau, : « A course on Adaptive Control »- 1
 Open Loop Adaptive Control of Deposited Zinc in
Hot-Dip Galvanizing
HOT DIP GALVANIZING (SOLLAC)
100% Digital Regulation
Standard Deviations 103%
. Computer aided
: 3.3%
manual control
. : 4.5%
% samples

% deposited zinc

Adaptation done with respect to:

• Steel strip speed 9 operation regions
• Distance between air knives and steel strip
The sampling period is tied to the strip speed to have constant discrete time delay 37
I.D. Landau, : « A course on Adaptive Control »- 1
 Direct Adaptive Control of a Phosphate Dryer Furnace

Large delay : 90 s
Better quality( reduction of the humidity standard deviation)
Reduction of fuel comsumption and of the thermal stress.
38
I.D. Landau, : « A course on Adaptive Control »- 1
Automatica 114 (2020) 108824

39
I.D. Landau, : « A course on Adaptive Control »- 1
 40
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control of a Flexible Transmission

The flexible transmission motor

axis load


m

axis
position Position
d.c. transducer
motor
D
y(t)
A
A u(t) C
Controller R-S-T
D
controller ref
C

41
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control of a Flexible Transmission
Frequency characteristics for various load

Rem.: the main vibration mode varies by 100%

Solution : Adaptive control with multiple models

42
I.D. Landau, : « A course on Adaptive Control »- 1
 Adaptive Control versus Robust Control

Load variations : 0%  100% (in 4 steps, 25% each)

Rem : The robust controller used is the winner of an international
benchmark test for robust control of the flexible transmission
(EJC, no.2., 1995)
43
I.D. Landau, : « A course on Adaptive Control »- 1
Rejection of unknown narrow band disturbances
in active vibration control

44
I.D. Landau, : « A course on Adaptive Control »- 1
 The
TheActive
ActiveSuspension
SuspensionSystem
System

machine primary acceleration / force (disturbance) Objective:

•Reject the effect of unknown
main elastomere cone
chamber 1
piston and variable narrow band
residual acceleration (force)
4 disturbances
hole
actuator
(piston controller
•Do not use an aditional
2
motor 3 position)
measurement

support inertia chamber

Two paths (Box Jenkins model):
•Primary
u p (t )
(disturbance)
•Secondary (double
-d1 differentiator)
q ⋅C / D
Controller Plant p1(t ) The model of the disturbance is
u(t ) -d + +
needed for the design
−
R/S q ⋅B / A y (t )
(residual force)
T s =1.25 m s
45
I.D. Landau, : « A course on Adaptive Control »- 1
 The Active Suspension

Active
suspension

Residual
force
(acceleration)
measurement

Primary force
(acceleration)
(the shaker)

46
I.D. Landau, : « A course on Adaptive Control »- 1
 Direct Adaptive Regulation : disturbance rejection
Disturbance : Chirp

25 Hz
47 Hz
Open
loop

Closed
loop

Initialization of the
adaptive controller
47
I.D. Landau, : « A course on Adaptive Control »- 1
 Direct Adaptive Regulation : rejection of sinusoidal disturbances
Simultaneous controller initialization Direct adaptive control
and disturbance application

Step changes in the frequency of the disturbance

48
I.D. Landau, : « A course on Adaptive Control »- 1
http://www.gipsa-lab.grenoble-inp.fr/~ioandore.landau/benchmark_adaptive_regulation/
49
I.D. Landau, : « A course on Adaptive Control »- 1
Adaptive Feedforward Compensation of Disturbances

50
I.D. Landau, : « A course on Adaptive Control »- 1
Feedforward compensation of disturbances. Why and How?

• Rejection (attenuation) of disturbances by feedback is limited

by the Bode integral*) of the output sensitivity function.

• To go beyond this limitation one has to use “feedforward”

compensation of disturbances.

• Feedforward compensation needs an additional transducer providing an

“image of the disturbance” (measurement correlated with the disturbance)

• For the design of the “feeforward compensator” the model of the disturbance
is needed.
|S yp|dB

*)For open loop stable systems, the integral of the logarithm 0 0.5fs
of the modulus of the output sensitivity function over the
frequency domain is 0
51
I.D. Landau, : « A course on Adaptive Control »- 1
An
Anactive
activevibration
vibrationcontrol
controlsystem
systemusing
usingan
aninertial
inertialactuator
actuator

Measurement of the image of the

disturbance y(t) Inertial actuator s(t)
Support (Disturbance source)
Global primary path: W(z-1)·D(z-1)
M1
Feedforward BM
compensator N Positive coupling (Reverse path): M(z-1) =
AM
M2
Parametric Adaptation BD
Algorithm Primary path: D(z-1) =
M3 AD
Residual acceleration χ(t)
Inertial actuator u(t)
Secondary path G(z-1) =B G / A G Support
(Compensator)
Power
Amplifier 52
I.D. Landau, : « A course on Adaptive Control »- 1
 IIR
IIRAdaptive
Adaptivefeedforward
feedforwarddisturbance
disturbancecompensation
compensation
Global primary path
s (t ) d (t ) x (t ) χ (t )
W D
Primary path
d (t ) Residual
Measurement of the image of the acceleration
(a) disturbance measurement
Global primary path

s (t) d (t ) x (t )
W D
Primary path 0
Positive feedback coupling
+ χ (t )
M + 1
Secondary path
+
N^ G
+ u^ ( t ) ^y ( t ) ^z 0 ( t )
Feedforward 0
filter PAA ν (t )
Measurement of the image
of the disturbance Parameter Adaptation Algorithm
(b)
53
I.D. Landau, : « A course on Adaptive Control »- 1
Adaptive feedforward compensation of a wide band disturbance

Without Effect of
compensation Power spectral density of the output compensation

54
I.D. Landau, : « A course on Adaptive Control »- 1
 References

ISBN: 1-84628-055-9
ISBN:978-0-85729-663-4
http://www.gipsa-lab.grenoble-inp.fr/~
ioandore.landau/identificationandcontrol/ http://www.landau-adaptivecontrol.org/55
I.D. Landau, : « A course on Adaptive Control »- 1
https://link.springer.com/book/10.1007%2F978-0-85729-664-1
 References
The course is mainly based on the following books:
• I.D. Landau, G. Zito “Digital control systems” Springer,London 2005
• I.D. Landau, R. Lozano, M. M’Saad, A. Karimi“Adaptive Control”,
Springer, London, 2011
• Y.D. Landau “Adaptive control – the Model reference approach” Dekker, NY, 1979
and papers:
• I.D. Landau “"Identification in closed loop- A powerful design tool
(better models, simpler controllers), Control Eng. Practice, no. 1, Jan. 2001
• I.D. Landau “From robust control to adaptive control” "Control Eng. Practice,
vol. 7,no10, pp1113-1124, 1999
• A. Karimi, I.D. Landau “Robust adaptive control of a flexible transmission system
using multiple models“, IEEE Trans. on CST, March 2000
• I.D. Landau, A. Constantinescu, D. Rey “Adaptive narrow band disturbance
rejection applied to an active suspension – an internal model approach”
Automatica, Vol. 41, n°4, 2005
• I.D. Landau, M. Alma, T.A. Airimitoaie “Adaptive feedforward compensation
algorithms for active vibration control with mechanical coupling” Automatica, Vol.47
pp. 2185-2196, 2011 I.D. Landau, : « A course on Adaptive Control »- 1
56
 Homework

57
I.D. Landau, : « A course on Adaptive Control »- 1