The Polynomial Toolbox for MATLAB

Index
Introduction The Polynomial Matrix Editor Polynomial matrix fractions Control system design Robust control with parametric uncertainties Numerical methods for polynomial matrices

Introduction

Index
Introduction The Polynomial Matrix Editor Polynomial matrix fractions Control system design Robust control with parametric uncertainties

The Polynomial Matrix Editor

The Polynomial Matrix Editor (PME) is recommended for creating and editing polynomial and standard MATLAB matrices of medium to large size, say from about 4 by 4 to 30 by 35 . Matrices of smaller size can easily be handled in the MATLAB command window with the help of monomial functions, overloaded concatenation, and various applications of subscripting and subassigning. On the other hand, opening a matrix larger than 30-by- 35 in the PME results in a window that is difficult to read.

Quick start Type pme to open the main window called Polynomial Matrix Editor. This window displays all polynomial matrices (POL objects) and all standard MATLAB matrices (2-dimensional DOUBLE arrays) that exist in the main MATLAB workspace. It also allows you to create a new polynomial or standard MATLAB matrix. In the Polynomial Matrix Editor window you can create a new polynomial matrix, by typing its name and size in the first (editable) line and then clicking the Open button modify an existing polynomial matrix while retaining its size and other properties: To do this just find the matrix name in the list and then double click the particular row.

 modify an existing polynomial matrix to a large extent (for instance by changing also its name, size, variable symbol, etc.): To do this, first find the matrix name in the list and then click on the corresponding row to move it up to the editable row. Next type in the new required properties and finally click Open . Each of these actions opens another window called Matrix Pad that serves for editing the particular matrix.

The main Polynomial Matrix Editor window is shown in Fig

Editable matrix

Opened editable box

Index
Introduction The Polynomial Matrix Editor Polynomial matrix fractions Control system design Robust control with parametric uncertainties

Polynomial matrix fractions

Linear time-invariant systems are a very important class of models for control. Even though the real world is without doubt thoroughly nonlinear, linear models provide an extraordinarily useful tool for the study of dynamical systems. A very well know model for linear time-invariant systems is of course the familiar state space description, which for continuous-time systems takes the form

For many applications the internal state or pseudo state x is not of interest, and only the external input and output variables u and y are relevant. It is not difficult to see that elimination of the internal variables by repeated differentiation and substitution in the continuous-time case leads to sets of differential equations in the output y and the input u that can be arranged in the form

State space systems and left polynomial matrix fractions

The Polynomial Toolbox command

[P,Q] = ss2lmf(A,B,C,D)
Q and P are left co prime and the transfer matrix Q-1 (s) P (s) equals the transfer matrix C( sI- A)-1 B+ D of the state space system. Q is row reduced and its row degrees are the observability indices of the state space system.

Example :from CT Chen book

Question:Find a right coprime fraction of the transfer matrix G(s)= [ (4s-10)/(2s+1) 3/(s+2); 1/(2s+1)(s+2) (s+1)/(s+2)^2] Answer: num = [ 4*s-10 3; 1 s+1 ];
den = [ 2*s+1 s+2; (2*s+1)*(s+2) (s+2)^2 ];

[Nl,Dl] = rat2rmf(num,den) Nl = -0.94 - 0.094s + 0.19s^2 -0.8 + 0.46s 0.047 0.11

Dl = 0.094 + 0.23s + 0.094s^2 0.11 + 0.23s 0 0.23 + 0.11s

[Nl,Dl] = rat2lmf(num,den) Nl = -0.9 - 0.09s + 0.18s^2 0.13 + 0.27s 0.4 - 0.13s 0.2
Dl = 0.09 + 0.22s + 0.09s^2 0 -0.066s 0.79 + 0.4s

lmf2rat, rmf2rat, rat2lmf, rat2rmf

In rational form the transfer matrix of a system with dimensions n by m is represented by two n* m polynomial matrices num and den. The entries of num are the numerators of the entries of the transfer matrix, and those of den the denominators. The commands lmf2rat, mf2rat, rat2lmf and rat2rmf provide conversion of polynomial matrix fractions to and from this format.

Index
Introduction The Polynomial Matrix Editor Polynomial matrix fractions Control system design Robust control with parametric uncertainties

Basic control routines

The Polynomial Toolbox offers basic functions to
stabilize the plant and, moreover, to parameterize all stabilizing controllers place closed-loop poles by dynamic output feedback

Stabilization:
A simple random stabilization can be achieved as follows. Given a linear time invariant plant with transfer matrix where v can be any of the variables s, p, z, q, z -1 or d , the command

[Nc,Dc] = stab(N,D)
computes a stabilizing controller with transfer matrix

The resulting closed-loop poles are randomly placed in the stability region, whose shape of course depends on the choice of the variable.

H-infinity

optimization

H-inf optimization is a powerful modern tool. It allows the design of high-performance and robust control systems. The Polynomial Toolbox offers two routines for H-inf design: Mixed sensitivity optimization of SISO systems relying on transfer function descriptions A routine dsshinf for finding all suboptimal solutions of the general standardH-inf optimization problem based on descriptor representations A very comprehensive routine dssrch for finding optimal solutions of the general standard H-inf optimization problem based on descriptor representations

Index
Introduction The Polynomial Matrix Editor Polynomial matrix fractions Control system design Robust control with parametric uncertainties

Introduction
Modern control theory addresses various problems involving uncertainty. A mathematical model of a system to be controlled typically includes uncertain quantities. In a large class of practical design problems the uncertainty may be attributed to certain coefficients of the plant transfer matrix. The uncertainty usually originates from physical parameters whose values are only specified within given bounds. An ideal solution to overcome the uncertainty is to find a robust controller a simple, fixed controller, designed off-line, which guarantees desired behavior and stability for all expected values of the uncertain parameters.

Single parameter uncertainty

Many systems of practical interest involve a single uncertain parameter. At the time of design the parameter is only known to lie within a given interval. Quite often even more complex problems (with a more complex uncertainty structure) may be reduced to the single parameter case. Needless to say that the strongest results are available for this simple case. Even though the uncertain parameter is single it may well appear in several coefficients of the transfer matrix at the same time.

Example:1
Steps to analyze this problem are as follow: 1. 2. Check whether p(s,q) is stable for q=0 Find left sided and right sided stability margins

With the Polynomial Toolbox this is an easy task: First express the given polynomial as

>> p0 = 3 + 10*s + 12*s^2 + 6*s^3 + s^4;

p1 = s + s^3; >> isstable(p0)

ans =

>> [qmin,qmax] = stabint(p0,p1) qmin = -5.6277 qmax = Inf

rlocus(ss(p1,p0),qmin:.1:100)

Example 2:Robust stabilization.

Steps for problem are as follow: 1. Suppose that q may take any value in the interval [0, 1] and that its nominal value is q0=0 2. The plant is described by a left-sided fraction of polynomial matrices in two variables: D(s,q) and N(s,q) that may be written as

And

Robust Control Structure

The closed loop denominator matrix is given by :

The denominator matrix may also be expressed as:

the matlab code is as follow:

D0 = [ s^2 1; 1 s ]; D1 = [ 0 1; 0 0 ]; D2 = [ 0 0; 1 0 ]; N0 = [ 1+s 0; 0 1 ]; N1 = [ 0 0; 1 0 ]; roots(D0) [Nc1,Dc1] = stab(N0,D0) P0 = D0*Dc1+N0*Nc1 P1 = D1*Dc1+N1*Nc1 P2 = D2*Dc1 roots(P0) [qmin,qmax] = stabint(P0,P1,P2)

[Nc2,Dc2] = stab(N0,D0) P0 = D0*Dc2+N0*Nc2 P1 = D1*Dc2+N1*Nc2 P2 = D2*Dc2 roots(P0) [qmin,qmax] = stabint(P0,P1,P2) qmin = -0.9344 qmax = 1.1700 Because the second controller evidently guarantees stability on the whole required uncertainty-bounding interval. Hence, it is the desired robustly stabilizing controller.

Interval polynomials
Another important class of uncertain systems is described by interval polynomials with independent uncertainties in the coefficients. An interval polynomial looks like

In many applications interval polynomials arise when an original uncertainty structure is known but too complex (e.g., highly nonlinear) to be tractable but may be overbounded by a simple interval once an independent uncertainty structure is imposed.

Zero Exclusion Principle Ref: page 38 from Robust control (the parametric approach) by S.P.Bhattacharya,Chapellat and Keel
Theorem1.6 Assume the family of polynomials is of constant degree,contains at least one stable polynomial and omega is path wise connected.then the entire family is stable iff 0 ( the family of polynomials).

Example3: Graphical Method

The first step in the graphical test for robust stability requires establishing that at least one polynomial in the family is stable. Using the midpoint of each of the intervals we obtain p_mid = pol([0.5 2 3 6 4 4 1],6) Matlab code: isstable(p_mid) ans=1 pminus = 0.45+1.95*s+2.95*s^2+5.95*s^3+3.95*s^4+3.95*s^5+s^6; pplus = 0.55+2.05*s+3.05*s^2+6.05*s^3+4.05*s^4+4.05*s^5+s^6; khplot(pminus,pplus,0:.001:1)

Since none of the rectangles touches the point z = 0 the Zero Exclusion Condition is satisfied. and we conclude that the interval polynomial is robustly stable.

Example 4:Test Using Kharitonov Polynomials.

For continuous-time interval polynomials we have an even simpler method available: An interval polynomial of invariant degree (with real coefficients) is known to be stable if and only if just its four extreme polynomials (called the Kharitonov polynomials)

are stable. For the interval polynomial of Example 3 the Kharitonov polynomials are computed by

[stability,K1,K2,K3,K4] = kharit(pminus,pplus)

Polytopes of polynomials
A more general class of systems is described by uncertain polynomials whose coefficients depend linearly on several parameters, but where each parameter may occur simultaneously in several coefficients. Such an uncertain polynomial may look like

with each coefficient ai(q) an affine function of q .

Uncertain polynomials with the affine uncertainty structure form polytopes in the space of polynomials. Similarly to the single parameter case such polynomials may always be expressed as

This form is preferred in the Polynomial Toolbox. Thus, a polytope of polynomials with n parameters is always described by the n + 1 polynomials p0( s ) , p1(s),. pn(s) along with n parameter bounding intervals

A simple calculation leads to the closed loop transfer function

If the plant has have an affine linear uncertainty structure then the closed-loop transfer function has an affine linear uncertainty structure as well.

we write

then the closed-loop characteristic polynomial follows as

while the numerator of the closed-loop transfer function is

Example 4:Improvement over rectangular bounds

we carry out two robust stability analyses. Part 1: Conservatism of Overbounding. First replace p( s,q) by the overbounding

Using the Kharitonov polynomials

pminus = pol([0.9 0.7 2.7 0.4 1],4); pplus = pol([ 4.6 1.3 8.3 1.6 1],4); [stable,K1,K2,K3,K4] = kharit(pminus,pplus); stable = 0
It is easy to verify that the third Kharitonov polynomial is unstable: isstable(K3) ans=0

Part 2: Value Set Comparison.

To begin with the second analysis, we express p( s, q) as

where

Matlab code :

p0 = pol([2 1 4 1 1],4);
p1 = pol([1 0 2],2); p2 = pol([-2 1 -1 2],3);

Qbounds = [-0.5 2; -0.3 0.3]

isstable(p0) ptopplot(p0,p1,p2,Qbounds,j*(0:0.025:2))

Summarizing, working with the overbounding interval polynomial is inconclusive while working with polygonal value sets leads us to the unequivocal conclusion that p (s, q) is robustly stable.

Example5:Robust stability degree design for a

polytopic plant. Consider the plant transfer function

with two uncertain parameters q1 [ 0, 0.2] and q2 [0,0.2] Both the numerator and the denominator of the transfer function are uncertain polynomials with a polytopic (affine) uncertainty structure. Write And

D0 = 2+s+2*s^2-2*s^3;
D1 = 1+2*s^2; D2 = -3*s;

N0 = 1+s;
N1 = 1; N2 = s; Qbounds = [ 0 0.2; 0 0.2 ] As the nominal plant

is unstable. Isstable(D0)

[Nc,Dc] = pplace(N0,D0,[-2,-2+j,-2-j,-3,-4]) The characteristic polynomial may be written as

Where P0 = D0*Dc+N0*Nc P1 = D1*Dc+N1*Nc P2 = D2*Dc+N2*Nc

ptopplot(10*P0,10*P1,10*P2,Qbounds,-.9+j*(0:.01:4))

The plot of Fig seems to indicate that zero is excluded. To be completely confident,we must zoom the picture to see the critical range 0<w<1. The closed-loop system is robustly stable.

Example 7.2 from book Robust control (the parametric approach) by S.P.Bhattacharya,Chapellat and Keel
G(s)=(s^3+ *s^2-2*s+ )/(s^4+2*s^3-s^2+*s+1) Matlab Code: D0 = [1-s^2+2*s^3+s^4]; D1 = [s]; N0 = [s^3-2*s]; N1 = [1] N2=[s^2] Qbounds = [ -1 -2;.5 1] [Nc,Dc] = pplace(N0,D0,[-2,-2+j,-2-j,-3,-4]) P0 = D0*Dc+N0*Nc P1 = D1*Dc+N1*Nc P2 =N2*Nc ptopplot(10*P0,10*P1,10*P2,Qbounds,-.9+j*(0:.01:4))

The plot of Fig seems to indicate that zero is excluded. To be completely confident,we must zoom the picture to see the critical range 0<w<1.The closed-loop system is robustly stable.

Questions ?