Scribd
Upload
127 views5 pages

OSC Sample Exam

The document provides details for three exam questions related to optimization methods in systems and control. Question 1 involves determining the best optimization method for 10 different problems and specifying an appropriate stopping criterion for each method. Question 2 involves proving properties about an optimization problem and showing equivalence between problems. Question 3 involves analyzing convexity of functions related to controller design problems.

Uploaded by

Remy Kabel
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
127 views5 pages

OSC Sample Exam

The document provides details for three exam questions related to optimization methods in systems and control. Question 1 involves determining the best optimization method for 10 different problems and specifying an appropriate stopping criterion for each method. Question 2 involves proving properties about an optimization problem and showing equivalence between problems. Question 3 involves analyzing convexity of functions related to controller design problems.

Uploaded by

Remy Kabel
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Sample Exam 1 Optimization in Systems and Control (SC4091)

QUESTION 1: Optimization methods (60 points)


Consider the following optimization problems: P1) P2) P3) P4) P5) P6) P7) P8) P9) P10)
xR2 xR2

min (x1 5)2 + (x2 8)4


xR3 x2

2 + x2 x1 2

1 25 10

min |x1 4 x2 + x3 4| ex3 z cos(z) + zx1 4 4 x3 z + 1


2 2 2

x dz

0 , x1 + x2

xR3

min

0 xR2 xR2 xR2

2 + log(x2 + 1) x2 1

2 2 x2 min ex1 +x2 + x1 2 + (x 10)2 x1 2 2 + (x 10)2 x1 2

2 4 2 + x2 ) + 14 x2 min cosh(x1

50 50

2 2 4 min cosh(x1 + x2 ) + 14 x2 3 min ex1 x2


2 2

xR2 xR2

x1 + x2 x1 + x2

15 0

min 25 x1 +2 x2 +6 x1 x2 +3 x1 x2

10 , x1 , x2 x x
1 2

2 2 x2 ) min ((x1 1)2 + (x2 2)2 )2 log(2 x1 2 2 x2 ) min (x1 + x2 )3 log(2 x1

1 1

xR2

a) Determine for each of the problems P1 up to P10 which of the following optimization methods is best suited: M1. Simplex method for linear programming. M2. Modied simplex method for quadratic programming. M3. Cutting-plane algorithm. M4. Davidon-Fletcher-Powell quasi-Newton algorithm. M5. Levenberg-Marquardt algorithm. M6. Perpendicular method with cubic line search. M7. Penalty function method with Nelder-Mead algorithm. M8. Gradient projection method with Fibonacci line search. M9. Sequential quadratic programming. Explain why the selected algorithm is the best suited for the given problem.

Note that you may have to perform some extra steps rst in order to simplify the optimization problem. If you do this, then you have to indicate it clearly. If it is necessary to use the selected algorithm in a multistart local minimization procedure, you have to indicate this clearly. b) Determine an appropriate stopping criterion for each of the problems P1 up to P10 and the corresponding selected optimization method. Explain your choice!

QUESTION 2: Optimization methods II (20 points)


a) Consider the following optimization problem:
x,yR

min f (x, y)

where f (x, y) = | 2x 2y | + | x + y 2 | . Prove that the function f is convex. Prove that (x, y) = (1, 1) is the global optimum of this optimization problem. Show that the perpendicular search method fails when starting at the initial point (x0 , y0 ) = (0, 0). b) Show that the condition y [0, 1] : with x, y R is satised if and only if

y1 x x y+1

1 x x 1 2 x x 2

0 0 .

Now consider the LMI-based optimization problem


,xR

min

1 x x 1 2 x x 2

0 0

and show that this optimization problem is equivalent to the optimization problem min max f (x, y)
xR y[0,1]

x2 . y+1 Hint: Use the Schur complement property for positive semi-denite matrices: if R( ) is invertible, then Q( ) S( ) 0 ST ( ) R( ) f (x, y) = y + 1 + 3

where

is equivalent to R( ) 0 , Q( ) S( )R1 ( )ST ( ) 0 .

QUESTION 3: Controller design using optimization (20 points)


a) Consider the system described by y(k) = a u(k) 1 q1

with a R and where q is the forward shift operator. The initial condition is y(0) = y0 . Let the input signal u be dened by u(k) = 1 for all k. Is y(k) convex as a function of a? Why? Let N be a positive integer and dene yav = Is yav convex as a function of a? Why? b) Consider the system described by y(k) = 1 u(k) 1 bq1 1 N y(k) . N k =1 (1)

with b R. The initial condition is y(0) = y0 . Let the input signal u be dened by u(k) = 1 for all k. Is y(k) convex as a function of b? Why? Is yav (cf. (1)) convex as a function of b? Why? c) Consider the system described by y(k) = d 1 bq1 cq2 u(k)

with b, c, d R. The initial condition is y(0) = y0 , y(1) = y1 . Let the input signal u be dened by u(k) = 1 for all k. Is y(k) convex as a function of b, c, and d ? Why? Is yav (cf. (1)) convex as a function of b, c, and d ? Why?

You might also like