Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Chair of Automatic Control

TUM School of Engineering and Design


Technical University of Munich

Note:
• During the attendance check a sticker containing a unique code will be put on this exam.
Eexam • This code contains a unique number that associates this exam with your registration
Place student sticker here number.
• This number is printed both next to the code and to the signature field in the attendance
check list.

Advanced Control
Exam: MW1420 / Module Exam Date: Thursday 18th August, 2022
Examiner: PD Dr.-Ing. habil. Paul Kotyczka Time: 13:45 – 15:15

Working instructions

• This exam consists of 24 pages with a total of 6 problems.


Please make sure on your own that you received a complete copy of the exam.
• The total amount of achievable credits in this exam is 90 credits.
• Detaching pages from the exam is prohibited.
• Allowed resources:

– One handwritten cheat sheet, DIN A4, double sided


– No calculators
– One analog dictionary English ↔ native language

• All answers must be given in the exam booklet. If the designated space is not sufficient, use the extra
pages inside the booklet and make a clear reference in the original field.

• For subproblems marked with *, the solutions of preceding questions are helpful or required.
• With the exception of single choice questions, the solution process or an explanation must be
given for all answers. Results are to be simplified and given without compound fractions.
• The correct solution for each problem must be clearly marked. Multiple solutions are not valid.

• Do not write with red or green colors nor use pencils.


• Physically turn off all electronic devices, put them into your bag and close the bag.
• Before the exam’s working time begins, there will be a reading time of 10 minutes during which writing
is prohibited.

Left room from to / Early submission at

– Page 1 / 24 –
Problem 1 Control of a SISO System (11 credits)
Consider the SISO system

α 2α − 1
   
0
ẋ = 0 1−α x+ 1 u (1.1)

y = [ 0 −2 ] x
with the parameter α ∈ R.

0 a) Give the eigenvalues of the system (1.1) and discuss the system’s asymptotic stability dependent on the
1 parameter α.
2

0 b) Use Kalman’s criterion to check the controllability of the system (1.1) dependent on the parameter α.
1
2
3

– Page 2 / 24 –
c) Now assume α = −1. Compute a state feedback u = −r T x using coefficient matching, such that the 0
eigenvalues of the closed-loop system are at −2 and −4. 1
2
3
4

Space for additional calculations.

– Page 3 / 24 –
For a certain value of α, the system (1.1) can be transformed into the diagonal form

−2 0 −1
   
ẋ = 0 3 x+ 1 u (1.2)

y = [ 2 0 ] x.

0 d) Use Gilbert’s criterion to check the observability of the eigenvalues of the diagonalized system (1.2).
1

0 e)* Is it possible to design an observer for the system (1.2), whose estimation error converges to zero?
1

Space for additional calculations.

– Page 4 / 24 –
Space for additional calculations.

– Page 5 / 24 –
Problem 2 Controller and Observer Design (17 credits)
Consider the general MIMO system

ẋ = Ax + Bu (2.1)
y = Cx,
which is to be controlled with the state feedback with constant feedforward gains

u = −Rx̂ + (RMx + Mu )w. (2.2)

Here, x̂ is the system state that is estimated by a full state observer. The observer is given by

x̂˙ = (A − LC)x̂ + Bu + Ly. (2.3)

0 a) Derive a general, extended state space model for the overall system, consisting of the plant (2.1) that is
1 controlled with the control law (2.2) and the observer (2.3).
2
3
4

– Page 6 / 24 –
b) The characteristic equation of the extended system (which consists of the controlled plant and the 0
observer) can be written 1
2
det(sI − A + BR) · det(sI − A + LC) = 0. (2.4)

Which statement can be made about the eigenvalues of the extended system based on (2.4)? What does
that mean for the assignment of eigenvalues when designing the state feedback controller and the observer?

c) How is the corresponding theorem called? 0


1

Space for additional calculations.

– Page 7 / 24 –
The general system (2.1) is now given by

−1
   
0 2 0
ẋ = 0 3 x+ 0 −1 u (2.5)

−1
 
1
y= 0 2 x.
The system can be considered fully controllable and observable.

0 d) The matrices Mx and Mu in the control law (2.2) must be calculated such that steady-state accuracy is
1 achieved, i.e. y = w holds in the steady state. Calculate Mx , if
2 " #
3
− 12 1
4
Mu = 3
0 2

has already been calculated.

– Page 8 / 24 –
e) Now calculate the feedback matrix R using parametric state feedback (Roppenecker’s formula). The 0
closed-loop eigenvalues shall be placed at λR1 = −2 and λR2 = −3. Use the parameter vectors 1
    2
2 0 3
p1 = 0 , p2 = −3. (2.6) 4
5
6

f) How does the feedback matrix R change, if the parameter vectors 0


1
−4
   
0
p1 = 0 , p2 = 1
are used instead of the vectors (2.6)? Give a short explanation (no calculations necessary).

– Page 9 / 24 –
Problem 3 Decoupling Control (19 credits)
The state space representation of a MIMO system is
   
1 2 0 1 0
ẋ =  0 2 1 x +  0 1 u (3.1)
1 0 0 0 0
 
a 1 0
y= 0 1 1 x,

where a ∈ R\{0} is a parameter of the system. For the given system, a controller

u = −Rx + F w (3.2)

is to be designed, which shall decouple the reference transfer behavior and ensure steady-state accuracy.
The poles of the reference transfer matrix Gw (s) of the resulting closed-loop system shall all be placed in
−4.

0 a) What is the relative degree of the system? Based on that, give the number of eigenvalues that can be
1 arbitrarily placed in the closed-loop system.
2
3
4

0 b) Show, that the system (3.1) can be decoupled.


1
2

– Page 10 / 24 –
c) Compute the invariant zero(s) of the system (3.1). 0
1
2
3

d)* For which range of values of a does the decoupling controller ensure stability? Provide a brief justification. 0
1
2

– Page 11 / 24 –
In the following, assume a = − 12 .

0 e) Compute the feedback matrix R and the prefilter F of the decoupling controller (3.2).
1
2
3
4
5
6

– Page 12 / 24 –
f)* Without explanation, give all resulting eigenvalues of the closed-loop system. 0
1

g) Give the reference transfer matrix Gw (s) of the closed-loop system. 0


1

Space for additional calculations.

– Page 13 / 24 –
Problem 4 Disturbance Rejection (18 credits)
Fig. 4.1 depicts the (normalized) block diagram of a DC machine with constant separate excitation. The
electrical armature circuit is approximated with a PT1 lag element with time constant T > 0. x2 is the
resulting motor torque, which is opposed by the load torque z . From the integrated resulting torque, multiplied
with a constant k > 0, which models the inverse moment of inertia, the angular velocity x1 is obtained, which
is also the system’s output. Due to the back electromagnetic force, which opposes the armature voltage u, a
feedback to the input arises.

u 0.1 x2 − k x1 = y
− Ts + 1 s
Armature circuit

Figure 4.1: Block diagram of the linearized DC machine.

0 a) From the block diagram 4.1, derive a complete state-space model of the DC machine in the form
1
2 ẋ = Ax + bu + ez
3
4 y = cT x
5
using the state vector x = [x1 , x2 ]T .

– Page 14 / 24 –
For certain values of T and k , the DC machine can be described with the system

0 3 0
     
3
ẋ = 1 x+ 1 u+ z (4.1)
− 2 −5 2
0
y = [ 1 0 ] x,
which will be considered from now on.
Furthermore, we assume that the disturbance z(t) can be described with the disturbance model

0 1
 
x̄˙ = 1 x̄ (4.2)
−4 0
z = [ 1 0 ] x̄.

b) Which type of disturbance does the disturbance model (4.2) describe (with x̄(0) 6= 0)? 0
1

c)* Explain, if the present disturbance (4.2) can be compensated (asymptotically) using a state feedback 0
controller with integral action. 1

Space for additional calculations.

– Page 15 / 24 –
0 d) For now, we assume that the disturbance is piecewise constant. Calculate a constant disturbance
1 feedback nx , nu for the system (4.1) that asymptotically compensates the effect of constant disturbances
2 on the output y .
3

Space for additional calculations.

– Page 16 / 24 –
e) In addition to the constant disturbance feedback, we now use an observer-based state feedback 0
controller and a constant feedforward controller. Assume, that the state x̄ of the disturbance model can 1
be measured (i.e., is not estimated by an observer). 2
3
Draw a block diagram of the closed-loop system containing the following components: 4
5
• A block of the plant, which is affected by the disturbance z(t) and labeled with „plant“. 6
7
• A block of the observer, which is labeled „observer“. 8

• A block of the disturbance model (for arbitrary disturbances), which is labeled „disturbance model“.
• The blocks R, Mx , Mu , Nx and Nu for the state feedback controller and the constant feedforward
and disturbance feedback.

Label all signals and summation points as known from the lecture.

– Page 17 / 24 –
Problem 5 Input-Output Linearization (9 credits)
We consider the so-called van der Pol oscillator, which is described by the nonlinear differential equations

ẋ1 = x2 (5.1)
ẋ2 = −x1 + (1 − x21 )x2 +u
with the system parameter  > 0.

0 a) Determine the relative degree of the system (5.1) for the two outputs
1
2 • yA = x1 and
3
4 • y B = x2 ,

respectively. Explain, which output is better suited for the design of an input-output-linearizing controller.

– Page 18 / 24 –
b)* Compute an input-output-linearizing controller u = f (x1 , x2 , w) for the van der Pol oscillator (5.1) 0
with y = x1 , which ensures linear, steady-state accurate input-output behavior. All assignable poles shall be 1
placed in −2. 2
3
4
5

– Page 19 / 24 –
Problem 6 Short Questions (16 credits)
Answer the following short questions by marking the correct answer. There is only one correct answer per
question. Each correct answer gives 2 points. Wrong answers or multiple marks per question give zero
points.

Mark correct answers with a cross ×


To undo a cross, completely fill out the answer option 
To re-mark an option, use a human-readable marking ×

a) Consider a real-valued matrix A ∈ R3×3 . Which constellation of eigenvalues is possible ?

−1j , +1j , −3.


−1, +1, −3j .
−1j , +1, −3.
−1j , +1j , −3j .

b) For the Kalman controllability matrix QC of a system

ẋ = Ax + Bu
y = Cx
with order n = 5, rank QC = 3 holds. The system does not have multiple eigenvalues. Which of the
following statements is wrong?

For at least one i ∈ {1, ..., 5}, it holds rank [λi I − AB] < 5.

There is exactly one invariant zero in the system.

The system is not fully controllable.

The system has two uncontrollable eigenvalues.

c) The task of the prefilter matrix F in the stabilizing control law u = −Rx + F w is to ensure steady-state
accuracy with respect to the reference value w . What must be satisfied, so that F can be computed?

The plant must not have an invariant zero with positive real part.

The plant must be asymptotically stable.

The plant must be stabilizable.

The plant must not have an invariant zero in zero.

d) With a state feedback u = −Rx + F w with F invertible, it is not possible to

make an observable system unobservable.

make a stable system unstable.

make a stabilizable system unstable.

make a controllable system uncontrollable.

– Page 20 / 24 –
e) Which of the following statements does not hold for Ackermann’s formula for the design of a state feedback
controller?

There exists a dual formula for observer design.

The plant that is to be controlled must have one input; the number of outputs is arbitrary.

Requirement is full controllability of the plant that is to be controlled.

Requirement is asymptotic stability of the plant that is to be controlled.

f) What are the design parameters for the calculation of an optimal controller / Riccati controller?

The symmetric, positive definite matrix P .

The weighting matrices Q and Q̃.

The value of the cost function J .

The closed-loop eigenvalues.

g) The dynamical system

ẋ = Ax + bu + ez
y = cT x
has exactly one invariant zero at η = 3. It is fully controllable and affected by a disturbance z . Is it possible
to achieve disturbance decoupling in the asymptotically stable closed-loop system using state feedback?

Yes, if one closed-loop eigenvalue λRi is shifted into the invariant zero, made unobservable and
e = kvRi with k 6= 0 holds, where vRi is the closed-loop eigenvector that corresponds to λRi .
Yes, if one closed-loop eigenvalue λRi is shifted into the invariant zero and made unobservable.

No, it is not possible to achieve disturbance decoupling and asymptotic stability in the closed-loop
system.

Yes, if there is at least one closed-loop eigenvalue λRi , for which e = kvRi with k 6= 0 holds, where
vRi is the closed-loop eigenvector that corresponds to λRi .

h) What is the advantage of using model-based dynamic feedforward control for disturbance rejection
to compensate piecewise constant disturbances?

The dynamics, with which the effects of the disturbances decline asymptotically, can be designed
independently from the plant dynamics.

The disturbance can be rejected completely (at all times).

It enables almost ideal reference tracking without any additional measures, i.e., y(t) ≈ w(t).

An unstable plant does not have to be stabilized with an additional state feedback controller.

– Page 21 / 24 –
Additional space for solutions–clearly mark the (sub)problem your answers are related to and strike
out invalid solutions.

– Page 22 / 24 –
– Page 23 / 24 –
– Page 24 / 24 –

You might also like