Organization

Time and Place: Wednesday, 9.4511.15 am, Seminarraum 3.243 Thursday, 9.4511.15 am, Seminarraum 3.243 Lecturers: Frank Allgwer, IST, PWR 9, room 2.246 [email protected] Carsten Scherer, IMNG, AR 5b, room 1.012 [email protected] Teaching assistants: Simone Schuler, PWR 9, room 2.235, [email protected] Martin Lhning, PWR 9, room 2.238, [email protected] Course webpage: www.ist.uni-stuttgart.de/education/courses/robust Exercises: 5 exercises plus miniprojects Course material: Mix of slides and handouts. For download at webpage.
Robust Control Part 1 p. 1

Textbooks
C. W. Scherer: Theory of Robust Control. (Textbook by course lecturer) S. Skogestad and I. Postlethwaite: Multivariable Feedback Control. (Theory, excellent explanations, many applications) K. Zhou, J. C. Doyle, and K. Glover: Robust and Optimal Control. (Lots of theoretical background)

Robust Control

Part 1 p. 2

Motivating Example 1
What you have done in Regelungstechnik I and II: Treated linear time-invariant systems and nonlinear systems Learned about classical controller design methods Consider a multi-input/multi-output (MIMO) system: Y1 (s) Y2 (s) = G(s) U1 (s) U2 (s) with G(s) =
1 s+1 0.5 s+1 4 s+8 1 s+1

Suppose we neglect off-diagonal terms and choose the control structure U1 (s) = K1 (s)(R1 (s) Y1 (s)) and U2 (s) = K2 (s)(R2 (s) Y2 (s)), i.e. U1 (s) U2 (s) = K(s) R1 (s) Y1 (s) R2 (s) Y2 (s) with K(s) = K1 (s) 0 0 K2 (s)

with K1 (s) = 20, K2 (s) = 18

Robust Control Part 1 p. 3

Motivating Example 1
K1 gives nice response for G11 , K2 gives nice response for G22 ,
Step response of closed SISO loops Gc1, Gc2 2

but: overall closed loop is unstable

Step response of overall closed loop Gc 10 8 6 4 2 0 2 4 6 8

Outputs y , y

1 0

Outputs y1, y2 1 2 Time 3 4 5

10 0

2 Time

Gc1 = G11 K1 /(1 + G11 K1 ) Gc2 = G22 K2 /(1 + G22 K2 )

Robust Control

Gc = GK(I + GK)1

Part 1 p. 4

Motivating Example 1
Why is the closed loop unstable? ; G has zero in right half-plane, and controllers gains are too big
1 s+1 0.5 s+1 4 s+8 1 s+1

G(s) =

Question 1: How can we determine properties of MIMO systems? Question 2: How should we design MIMO controllers for MIMO plants?

Robust Control

Part 1 p. 5

Motivating Example 1
Two different responses of closed loop:
Step response of overall closed loop Gc 2 Step response of overall closed loop Gc 2

Outputs y , y

Outputs y1, y2 1 2 Time 3 4 5

1 0

1 0

2 Time

Question 3: How can we determine performance quantitatively? Question 4: How to design controllers that guarantee good performance?
Robust Control Part 1 p. 6

Motivating Example 2
Bode magnitude plot of plant models 20

Real system: Greal (s) =

Magnitude

1 1 s + 1 (0.1s + 1)2

20 40 60 80 2 10 Gr G 10 Frequency
0

Approximation: G(s) = 1 s+1

10

Idea: Bode plots are similar, differences only for large freqs. Place poles far from stability border in left half-plane to be on safe side

Robust Control

Part 1 p. 7

Motivating Example 2
Controller: 30(s + 1) K(s) = s achieves Gcl (s) = GK 30 = 1 + GK s + 30
Output 3 2 1 0 1 2 with G with G 0.5
r

Step response of closed loops

i.e. pole at s = 30

but simulation of real closed loop shows instability!

3 0

1 Time

1.5

Question 5: What means good robustness / to be on safe side? Question 6: How to analyze robustness? Question 7: How to achieve robustness systematically?
Robust Control Part 1 p. 8

Motivating Example 3
Consider a simple train model with varying load:

m = F Ffriction x Mass m varies between 100 t and 250 t Friction force Ffriction is an unknown disturbance

Question 8: How to suppress the disturbance? Question 9: How to design a robust controller, i.e. a controller that achieves stability and good performance for all values of m?

Robust Control

Part 1 p. 9

Summary of Motivation
The Robust Control Course treats analysis of stability, analysis of performance specications, and synthesis of suitable controllers for multivariable systems with uncertainties.

Already about a hundred years ago, German poet Joachim Ringelnatz (18831934) recognized: Sicher ist, dass nichts sicher ist. Selbst das nicht! The only certain thing is that nothing is certain. Not even that!

Robust Control

Part 1 p. 10

Lecture Overview
1 Multi-input/multi-output systems
Description and properties Stability and performance analysis

Uncertain systems
Description Robust stability analysis Robust performance analysis

Controller design methods

Review of classical design methods H control synthesis

Robust Control

Part 1 p. 11

Course Announcement

Short course on Asymptotic Tracking and Disturbance Rejection

Prof. Alberto Isidori
Universit di Roma La Sapienza Rome, Italy

Course Announcement

Short Course on Asymptotic Tracking and Disturbance Rejection

Organizational Information
Time and Place
May 3rd, 5th, and 7th each 9:45-11:30 and 14:00-15:30 Seminar room of the IST (Pfaffenwaldring 9, 3.243). 1 SWS, e.g. Systemtheorie (kyb) or Hauptfach (mach,. . . ) Required: RT I and RT II or equivalent lectures.

More Information
Course homepage http://www.ist.uni-stuttgart.de courses Short Course IST notice board Assistant: Marcus Reble [email protected]
Course Announcement Short Course on Asymptotic Tracking and Disturbance Rejection 2

Part 1: MIMO LTI Systems

Description of linear time-invariant systems Conversion between different descriptions Operations on systems More general system classes: linear time-varying, linear parameter-varying, nonlinear Continuous-time versus discrete-time systems

Robust Control

Part 1 p. 12

Linear Time-Invariant (LTI) Systems

State-space form x(t) = Ax(t) + Bu(t), y(t) = Cx(t) + Du(t) x(t0 ) = x0

with x(t) Rn , y(t) Rp , u(t) Rq often short x = Ax + Bu, y = Cx + Du Response to input signal u : [0, ) Rq
t

or

{A, B, C, D}

x(t) = e

A(tt0 )

x0 +
t0

eA(t ) Bu( )d
t

y(t) = Ce

A(tt0 )

x0 +
t0

CeA(t ) Bu( )d + Du(t)

Assume all signals to be piecewise continuous Similarity transformation with x = T x, leads to

Robust Control

T non-singular (i.e. det T = 0) y = CT x + Du, x0 = x(t0 ) = T 1 x0

Part 1 p. 13

x = T 1 AT x + T 1 Bu,

has same input/output (I/O) behavior, but states have different meaning

Linear Time-Invariant (LTI) Systems

What is the systems I/O behavior?
t

set t0 = 0, x0 = 0

unique output signal

y(t) =
0

CeA(t ) Bu( )d + Du(t)

Dene Laplace transformation Then I/O behavior is

X(s) = L(x(t)) :=
0

x(t)est dt

Y (s) = (C(sI A)1 B + D )U (s)

G(s)

G(s) is called transfer matrix:

nij (s) dij (s)

where

Gij (s) =

with nij , dij real polynomials in s

G11 (s) . . G(s) = . Gp1 (s)

G1q (s) . , . . Gpq (s)

G is called proper if lim G(s) = const.,

s

i.e. deg(nij ) deg(dij ) for all i, j i.e. deg(nij ) < deg(dij ) for all i, j
Part 1 p. 14

G is called strictly proper if lim G(s) = 0,

s
Robust Control

Linear Time-Invariant (LTI) Systems

Use notation C D to represent state-space realization of transfer matrix G(s) = A B := C(sI A)1 B + D

Impulse matrix

g(t) = L1 (G(s)) = CeAt Bh(t) + D(t),

where h(t) = unit step, (t) = unit impulse Then I/O behavior by convolution:

y(t) = (g u)(t) =
0

g(t )u( )d =
0

g( )u(t )d

Robust Control

Part 1 p. 15

Linear Time-Invariant (LTI) Systems

Some special forms of transfer matrix representations: Polynomial matrix fraction representation (MFR): Right polynomial MFR: G(s) = N (s)M (s)1 Left polynomial MFR: G(s) = M (s)1 N (s) where N , M , M , N are polynomial matrices in s Coprime MFR over RH : Right coprime MFR: G(s) = N (s)M (s)1 Left coprime MFR: G(s) = M (s)1 N (s) where N , M / M , N are right / left coprime matrices with proper, asymptotically stable, real rational components, i.e. N , M , N , M RH

Robust Control

Part 1 p. 16

Operations on Systems
G(s) = A C B D = C(sI A)1 B + D
T

Transpose:

G(s) =

AT B
T

CT D
T

also called dual of G

Conjugate system:

G(s) = G(s) =

AT B B
T

C T D A C
T

AT B T

CT DT

Negated system:

G(s) =

A C

B D

Inverse:

if G(s) square (p = q) and det(D) = 0, then G(s)

1

A BD 1 C D 1 C

BD 1 D 1
Part 1 p. 17

Robust Control

Operations on Systems
G1 (s) = A1 C1 B1 D1 , G2 (s) = A2 C2 B2 D2 A1 G1 (s) + G2 (s) = 0 C1 A1 G1 (s) G2 (s) = 0 C1 A 1 0 = C1 0 0 A2 0 C2 0 A2 C2 B1 C2 A2 D1 C2 B1 B1 B2 D1 + D2 B1 D 2 B2 D1 D2

Addition / parallel connection:

Multiplication / series connection:

Stacking:

G1 (s) G2 (s)

B2 D1 D2
Part 1 p. 18

Robust Control

Operations on Systems
G1 G2
Feedback interconnection: G(s) = G1 (I + G2 G1 )1 = (I + G1 G2 )1 G1 1 1 A1 B1 D2 V12 C1 B1 V21 C2 1 1 = A2 B2 D1 V21 C2 B2 V12 C1 1 1 V12 C1 V12 D1 C2 where V12 := I + D1 D2 and V21 := I + D2 D1 . Last four rules may result in non-minimal realizations, even if {A1 , B1 , C1 , D1 } and {A2 , B2 , C2 , D2 } are minimal.

1 B1 V21 1 B2 D1 V21 1 V12 D1

Robust Control

Part 1 p. 19

Further System Classes

Time delay systems: here input delay (similarly for state delay, output delay) x(t) = Ax(t) + Bu(t ) y(t) = Cx(t) + Du(t ) Linear time-varying (LTV) systems: x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t) Linear parameter-varying (LPV) systems: x(t) = A(p(t))x(t) + B(p(t))u(t) y(t) = C(p(t))x(t) + D(p(t))u(t) sometimes p() is measurable
Robust Control Part 1 p. 20

Further System Classes

Input-afne non-linear systems: x(t) = f (x(t)) + g(x(t))u(t) y(t) = h(x(t)) Non-linear systems: x(t) = f (x(t), u(t)) y(t) = h(x(t), u(t)) Often possible to view a system from perspective of different classes Often possible to treat these system classes with robust control methods (possibly with conservatism)

Robust Control

Part 1 p. 21