Solutions 1-4 by Poya Khalaf

Similar documents
Systems Analysis and Control

Outline. Classical Control. Lecture 5

Modern Control Systems

Mathematics for Control Theory

Problem Set 5 Solutions 1

ECE 3793 Matlab Project 3 Solution

Homework 11 Solution - AME 30315, Spring 2015

Asynchronous Training in Wireless Sensor Networks

E2.5 Signals & Linear Systems. Tutorial Sheet 1 Introduction to Signals & Systems (Lectures 1 & 2)

D(s) G(s) A control system design definition

FRTN 15 Predictive Control

GATE EE Topic wise Questions SIGNALS & SYSTEMS

ECEN 420 LINEAR CONTROL SYSTEMS. Lecture 2 Laplace Transform I 1/52

EE/ME/AE324: Dynamical Systems. Chapter 7: Transform Solutions of Linear Models

6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski. Solutions to Problem Set 1 1. Massachusetts Institute of Technology

Control Systems I. Lecture 6: Poles and Zeros. Readings: Emilio Frazzoli. Institute for Dynamic Systems and Control D-MAVT ETH Zürich

Advanced Control Theory

Bangladesh University of Engineering and Technology. EEE 402: Control System I Laboratory

Introduction. Performance and Robustness (Chapter 1) Advanced Control Systems Spring / 31

Solutions to Homework 3

Section Kamen and Heck And Harman. Fourier Transform

Analysis and Design of Control Systems in the Time Domain

Input-output Controllability Analysis

Exercise 3: Transfer functions (Solutions)

Frequency methods for the analysis of feedback systems. Lecture 6. Loop analysis of feedback systems. Nyquist approach to study stability

Chapter 6 - Solved Problems

EE 3CL4: Introduction to Control Systems Lab 4: Lead Compensation

Problem Set #7 Solutions Due: Friday June 1st, 2018 at 5 PM.

Goodwin, Graebe, Salgado, Prentice Hall Chapter 11. Chapter 11. Dealing with Constraints

Step Response Analysis. Frequency Response, Relation Between Model Descriptions

Problem Value

L2 gains and system approximation quality 1

(i) Represent continuous-time periodic signals using Fourier series

2 Background: Fourier Series Analysis and Synthesis

These videos and handouts are supplemental documents of paper X. Li, Z. Huang. An Inverted Classroom Approach to Educate MATLAB in Chemical Process

Loop shaping exercise

Lecture 8 ELE 301: Signals and Systems

Exercise 8: Level Control and PID Tuning. CHEM-E7140 Process Automation

6.241 Dynamic Systems and Control

Lecture 8. Chapter 5: Input-Output Stability Chapter 6: Passivity Chapter 14: Passivity-Based Control. Eugenio Schuster.

16.31 Homework 2 Solution

Question Paper Code : AEC11T02

Problem Value

Control Systems Lab - SC4070 Control techniques

Lab-Report Control Engineering. Proportional Control of a Liquid Level System

Problem Set 2 Solutions 1

Cosc 3451 Signals and Systems. What is a system? Systems Terminology and Properties of Systems

Review of Frequency Domain Fourier Series: Continuous periodic frequency components

Raktim Bhattacharya. . AERO 422: Active Controls for Aerospace Vehicles. Dynamic Response

Laboratory 1. Solving differential equations with nonzero initial conditions

Lecture 4&5 MATLAB applications In Signal Processing. Dr. Bedir Yousif

Numerical Solutions to Partial Differential Equations

Lab Experiment 2: Performance of First order and second order systems

This homework will not be collected or graded. It is intended to help you practice for the final exam. Solutions will be posted.

CHAPTER 6 STATE SPACE: FREQUENCY RESPONSE, TIME DOMAIN

Non-homogeneous equations (Sect. 3.6).

The Laplace Transform

On linear and non-linear equations. (Sect. 1.6).

Automatic Control III (Reglerteknik III) fall Nonlinear systems, Part 3

MAT 275 Laboratory 7 Laplace Transform and the Symbolic Math Toolbox

Theory of Linear Systems Exercises. Luigi Palopoli and Daniele Fontanelli

Linear Filters. L[e iωt ] = 2π ĥ(ω)eiωt. Proof: Let L[e iωt ] = ẽ ω (t). Because L is time-invariant, we have that. L[e iω(t a) ] = ẽ ω (t a).

Massachusetts Institute of Technology

Discrete-Time Fourier Transform

Deakin Research Online

H 2 Optimal State Feedback Control Synthesis. Raktim Bhattacharya Aerospace Engineering, Texas A&M University

Laplace Transforms and use in Automatic Control

CONTROL OF DIGITAL SYSTEMS

MAE143A Signals & Systems - Homework 5, Winter 2013 due by the end of class Tuesday February 12, 2013.

06/12/ rws/jMc- modif SuFY10 (MPF) - Textbook Section IX 1

EE C128 / ME C134 Fall 2014 HW 6.2 Solutions. HW 6.2 Solutions

100 (s + 10) (s + 100) e 0.5s. s 100 (s + 10) (s + 100). G(s) =

7. Find the Fourier transform of f (t)=2 cos(2π t)[u (t) u(t 1)]. 8. (a) Show that a periodic signal with exponential Fourier series f (t)= δ (ω nω 0

ECE 3793 Matlab Project 3

CDS 101/110: Lecture 3.1 Linear Systems

x(t) = t[u(t 1) u(t 2)] + 1[u(t 2) u(t 3)]

1 The Observability Canonical Form

EE 341 A Few Review Problems Fall 2018

RELAY CONTROL WITH PARALLEL COMPENSATOR FOR NONMINIMUM PHASE PLANTS. Ryszard Gessing

EEE 303 Notes: System properties

Course Summary. The course cannot be summarized in one lecture.

CH.6 Laplace Transform

Introduction to Modern Control MT 2016

Modeling and Analysis of Systems Lecture #8 - Transfer Function. Guillaume Drion Academic year

Meeting Design Specs using Root Locus

DSC HW 4: Assigned 7/9/11, Due 7/18/12 Page 1

Review of Fourier Transform

Matlab Controller Design. 1. Control system toolbox 2. Functions for model analysis 3. Linear system simulation 4. Biochemical reactor linearization

MATHEMATICAL MODELING OF CONTROL SYSTEMS

Problem Value

Continuous-Time Fourier Transform

Chapter 3 Convolution Representation

Problem set 5 solutions 1

Math Fall Linear Filters

Process Control Exercise 2

Control Systems II. ETH, MAVT, IDSC, Lecture 4 17/03/2017. G. Ducard

The Laplace Transform

The z-transform Part 2

Numerical Solutions to Partial Differential Equations

Digital Control System Models. M. Sami Fadali Professor of Electrical Engineering University of Nevada

Transcription:

Solutions -4 by Poya Khalaf. Consider the function: f { t < [ e t sin(t), te 2t] T t Calculate f 2,[, ] using the time-domain definition and then using Parsevals identity. The following code has been written to calculate the integrals using Matlab s symbolic toolbox: clc 2 clear 3 4 syms t w T 5 6 f=[exp(-t)*sin(t),t*exp(-2*t)].'; 7 8 %time domain definition 9 % int ˆinf [ f ˆ2dt disp('time domain definition') intf=int(f()ˆ2+f(2)ˆ2,t,,inf); 2 vpa(intf) 3 %frequency domain definition 4 %Fourier transform 5 F(,)=int(f().*exp(-i*w*t),t,,inf); 6 F(2,)=int(f(2).*exp(-i*w*t),t,,inf); 7 8 %int -infˆinf F*Fdw 9 intf=/(2*pi)*int(f'*f,w,-inf,inf); 2 disp('frequency domain definition') 2 vpa(intf,5) The results are as follows: time domain definition 2 ans = 3.5625 4 frequency domain definition 5 ans = 6.5625 2. Find an example of two LTI systems (to be cascaded) where the H 2 norm violates the submultiplicative property. Consider the two systems G = iω+ and H = 2 i ω+. The following code has been wriiten to calulate the 2-norm for the cascade system GH and G 2 H 2 : syms w 2 G=/(i*w+); 3 4 H=/(2*i*w+); 5 6 norm2gh=(/(2*pi)*int((g*h)'*(g*h),-inf,inf))ˆ.5; 7 disp(' GH ') 8 vpa(norm2gh,5) 9 norm2g=(/(2*pi)*int((g)'*(g),-inf,inf))ˆ.5; norm2h=(/(2*pi)*int((h)'*(h),-inf,inf))ˆ.5; 2 3 disp(' G * H ') 4 vpa(norm2g*norm2h,5)

The results are as follows: GH 2 ans = 3.4825 4 G * H 5 ans = 6.35355 It is seen that the 2-norm does not satisfy the submultiplicative property. Consider a system G(s) formed by cascading a pure time-shift operator S(s) and a first-order lag L(s): G(s) = L(s)S(s) S(s) = e bs L(s) = τs + Recall that the time-shift operator has the input-output relationship: u(t) u(t + b). When b <, this represents a delay, and b > corresponds to an advance. Pick your own value of τ. In this exercise, you will be illustrating the definition of causality in relation to G(s). a) Choose a value of b < and a truncation time T > to facilitate your example. Suppose the input to G(s) is a unit step. Use sketches of the signals at various points to Show that P T GP T u = P T Gu in this case. b) Now use a positive value of b with the same absolute value as above and the same T. Use sketches to show that P T GP T u P T Gu. c) The above proves (by counterexample) that G(s) is not causal if b >. For extra credit, prove mathematically that G is causal if and only if b <. a) For this purpose the following Matlab code has been written: s=tf('s'); 2 tau=.5; 3 T=3; 4 b=-; 5 6 L=/(tau*s+); 7 8 % Lu 9 [y,t]=step(l); t2=t; plot(t,y) 2 xlabel('time(s)') 3 ylabel('amplitude') 4 5 % Gu 6 t=t-b; 7 hold on 8 plot(t,y,'r') 9 title(['b=',num2str(b)]) 2 2 %P TGu 22 y(t>t)=; 23 hold on 24 plot(t,y,'k') 2

25 legend('lu','gu','p TGu') 26 27 % P Tu 28 u=ones(length(t2),); 29 u(t2>t)=; 3 3 %GP Tu 32 [y2,t2]=lsim(l,u,t2); 33 t2=t2-b; 34 figure 35 plot(t2,y2) 36 xlabel('time(s)') 37 ylabel('amplitude') 38 title(['b=',num2str(b)]) 39 4 %P TGP Tu 4 y2(t2>t)=; 42 hold on 43 plot(t2,y2,'k') 44 legend('gp Tu','P TGP Tu') 45 46 47 figure 48 plot(t,y,t2,y2) 49 legend('p TGu','P TGP Tu') 5 xlabel('time(s)') 5 ylabel('amplitude') 52 title(['b=',num2str(b)]) The results are as follows:.9.8.7 b= Lu Gu P T Gu Amplitude.6.5.4.3.2. 2 3 4 5 6 time(s) 3

.9.8 b= GP T u P T GP T u.7 Amplitude.6.5.4.3.2..5 2 2.5 3 3.5 4 4.5 5 5.5 time(s).9.8 b= P T Gu P T GP T u.7 Amplitude.6.5.4.3.2..5 2 2.5 3 3.5 4 4.5 5 5.5 time(s) b) Using the above code for b = : 4

.9.8.7 b= Lu Gu P T Gu Amplitude.6.5.4.3.2. 2 3 4 5 time(s).9.8 b= GP T u P T GP T u.7 Amplitude.6.5.4.3.2..5.5.5 2 2.5 3 3.5 time(s) 5

.9.8 b= P T Gu P T GP T u.7 Amplitude.6.5.4.3.2..5.5.5 2 2.5 3 3.5 time(s) c) Initially we find the expression for P T Gu. The transfer function L in the time domain is equal to: ẏ = τ y + τ u Also G = LS. In order to calculate GU we intially calculate SU. In the time domain SU is equal to: SU(t) = u(t + b) GU is found to be: Finally, P T GU is equal to: GU(t) = e t τ y + e t τ t e z τ u(z + b)dz Next, we calculate P T GP T U. P T U is equal to: { e t τ y P T GU(t) = + e t t τ e z τ u(z + b)dz t T t > T P T U(t) = { u t T t > T SP T U is found to be: GP T U is equal to: GP T U(t) = SP T U(t) = { { u(t + b) t T b t > T b e t τ y + e t t τ e z τ u(z + b)dz t T b e t τ y + e t T b τ e z τ u(z + b)dz t > T b 6

Now to calculate P T GP T U we consider two cases, b > and b <. In the first case if b > then T b < T and we have: e t τ y + e t t τ e z τ u(z + b)dz t T b P T GP T U(t) = e t τ y + e t T b τ e z τ u(z + b)dz T b < t < T t > T In this case we see that P T GP T U P T Gu. For the case b <, T b = T + b > T. Therefore P T GP T is found to be: { e t τ y P T GP T U(t) = + e t t τ e z τ u(z + b)dz t T t > T In this case we see that P T GP T U = P T Gu. Refer to the proof of the Small Gain Theorem in Green and Limebeer, Sect. 3.4. Provide full justification for the derivation of the inequality Se 2T Sê 2T 2,[,T ] γ(g )γ(g 2 ) e 2T ê 2T 2,[,T ] From the definition of Se 2T we have: Se 2T Sê 2T 2,[,T ] = G (w T + P T (G 2 e 2T )) G (w T + P T (G 2 ê 2T )) 2,[,T ] Since G has finite incremental gain we can write: G (w T + P T (G 2 e 2T )) G (w T + P T (G 2 ê 2T )) 2,[,T ] γ(g ) w T +P T (G 2 e 2T ) w T P T (G 2 ê 2T ) 2,[,T ] Which simplifies to: Se 2T Sê 2T 2,[,T ] γ(g ) P T (G 2 e 2T ) P T (G 2 ê 2T ) 2,[,T ] From the definition of the truncation operator we have: Since the norm is on [, T ] we can write: Se 2T Sê 2T 2,[,T ] γ(g ) P T (G 2 e 2T G 2 ê 2T ) 2,[,T ] Now since G 2 has finite incremental gain we have: Se 2T Sê 2T 2,[,T ] γ(g ) G 2 e 2T G 2 ê 2T 2,[,T ] Se 2T Sê 2T 2,[,T ] γ(g ) G 2 e 2T G 2 ê 2T 2,[,T ] γ(g )γ(g 2 ) e 2T ê 2T 2,[,T ] And therefore: Se 2T Sê 2T 2,[,T ] γ(g )γ(g 2 ) e 2T ê 2T 2,[,T ] 7

ESC794 Simulation Results: HWG The sliding mode controller was used first strictly as an output feedback controller (no feedback from signal x behind the saturation block). For all simulations λ was chosen as, and the switching gain η was chosen as. Figure shows the simulation diagram for this case. A saturation level M = was used first and the sign function in the controller was approximated using a saturation function applied to.s. The controller is able to arrest exponential growth of x and bring all internal signals to zero. The results are shown in Fig. 2. Note that the lower saturation block is active in several intervals. However,.s <, so the SMC is acting linearly. This is essentially a PI controller. Next, if M is reduced to.5, this controller is unable to stabilize x, as shown in Fig. 3. As a result, e and e 2 are not in L 2. Adding information about x as feedback will improve the controller, although this is no longer an output-based solution. As explained in the analysis, the term (λ+)x should be used in the control law. This is shown in the simulation diagram of Fig. 4. The controller is re-tuned to use s directly as the input to the saturation block. This time, M can be set to a value as low as. with internal stability. The results are shown in Fig. 5. A formal proof of stability would require advance knowledge of bounds on the external signals u and u 2, as well as the saturation level M.

s Step To Workspace7 s. Integrator Gain Gain3 Saturation Gain2 Pulse Generator Product WITHOUT x STATE FEEDBACK Gain u e To Workspace2 y To Workspace6 To Workspace3 x e2 To Workspace4 t To Workspace5 Clock To Workspace Saturation s Transfer Fcn Product u2 Pulse Generator To Workspace Figure : Simulation diagram of output-feedback SMC Results with saturation level M=, no X feedback 6 Signals x and y 5-5 - x y Sliding Function s 4 2-2 -4-5 2 4 6 8-6 2 4 6 8.5 6 4 e.5 e 2 2-2 -.5-2 4 6 8 Time -4-6 2 4 6 8 Time Figure 2: Results with output-feedback SMC and M =

5 Results with saturation level M=.5, no X feedback Signals x and y - -2-3 x y Sliding Function s - -2-3 -4 2 4 6 8-4 2 4 6 8 5.5 5 e e 2-5 -.5 2 4 6 8-2 4 6 8 Time Time Figure 3: Results with output-feedback SMC and M =.5 s Step To Workspace7 s Integrator Gain Gain3 Saturation Gain2 Pulse Generator Product WITH x STATE FEEDBACK Gain u e To Workspace2 y To Workspace6 To Workspace3 x e2 To Workspace4 t To Workspace5 Clock To Workspace Saturation s Transfer Fcn Product u2 Pulse Generator To Workspace Figure 4: Simulation diagram of SMC with x feedback

Signals x and y -2-4 -6-8 Results with saturation level M=. and X feedback 2 x y Sliding Function s 5 5-2 3 4-5 2 3 4.8.6 2-2 e.4 e 2-4.2 -.2 2 3 4 Time -6-8 - 2 3 4 Time Figure 5: Results with x feedback SMC and M =.