National Instruments Computer Accessories 370757C 01 User Manual

TM  
NI MATRIXx  
TM  
Xmath Robust Control Module  
MATRIXx Xmath Robust Control Module  
April 2007  
370757C-01  
 
Important Information  
Warranty  
The media on which you receive National Instruments software are warranted not to fail to execute programming instructions, due to defects  
in materials and workmanship, for a period of 90 days from date of shipment, as evidenced by receipts or other documentation. National  
Instruments will, at its option, repair or replace software media that do not execute programming instructions if National Instruments receives  
notice of such defects during the warranty period. National Instruments does not warrant that the operation of the software shall be  
uninterrupted or error free.  
A Return Material Authorization (RMA) number must be obtained from the factory and clearly marked on the outside of the package before any  
equipment will be accepted for warranty work. National Instruments will pay the shipping costs of returning to the owner parts which are covered by  
warranty.  
National Instruments believes that the information in this document is accurate. The document has been carefully reviewed for technical accuracy. In  
the event that technical or typographical errors exist, National Instruments reserves the right to make changes to subsequent editions of this document  
without prior notice to holders of this edition. The reader should consult National Instruments if errors are suspected. In no event shall National  
Instruments be liable for any damages arising out of or related to this document or the information contained in it.  
EXCEPT AS SPECIFIED HEREIN, NATIONAL INSTRUMENTS MAKES NO WARRANTIES, EXPRESS OR IMPLIED, AND SPECIFICALLY DISCLAIMS ANY WARRANTY OF  
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. CUSTOMERS RIGHT TO RECOVER DAMAGES CAUSED BY FAULT OR NEGLIGENCE ON THE PART OF NATIONAL  
INSTRUMENTS SHALL BE LIMITED TO THE AMOUNT THERETOFORE PAID BY THE CUSTOMER. NATIONAL INSTRUMENTS WILL NOT BE LIABLE FOR DAMAGES RESULTING  
FROM LOSS OF DATA, PROFITS, USE OF PRODUCTS, OR INCIDENTAL OR CONSEQUENTIAL DAMAGES, EVEN IF ADVISED OF THE POSSIBILITY THEREOF. This limitation of  
the liability of National Instruments will apply regardless of the form of action, whether in contract or tort, including negligence. Any action against  
National Instruments must be brought within one year after the cause of action accrues. National Instruments shall not be liable for any delay in  
performance due to causes beyond its reasonable control. The warranty provided herein does not cover damages, defects, malfunctions, or service  
failures caused by owner’s failure to follow the National Instruments installation, operation, or maintenance instructions; owner’s modification of the  
product; owner’s abuse, misuse, or negligent acts; and power failure or surges, fire, flood, accident, actions of third parties, or other events outside  
reasonable control.  
Copyright  
Under the copyright laws, this publication may not be reproduced or transmitted in any form, electronic or mechanical, including photocopying,  
recording, storing in an information retrieval system, or translating, in whole or in part, without the prior written consent of National  
Instruments Corporation.  
National Instruments respects the intellectual property of others, and we ask our users to do the same. NI software is protected by copyright and other  
intellectual property laws. Where NI software may be used to reproduce software or other materials belonging to others, you may use NI software only  
to reproduce materials that you may reproduce in accordance with the terms of any applicable license or other legal restriction.  
Trademarks  
MATRIXx, National Instruments, NI, ni.com, and Xmathare trademarks of National Instruments Corporation. Refer to the Terms of  
Use section on ni.com/legalfor more information about National Instruments trademarks.  
Other product and company names mentioned herein are trademarks or trade names of their respective companies.  
Members of the National Instruments Alliance Partner Program are business entities independent from National Instruments and have no agency,  
partnership, or joint-venture relationship with National Instruments.  
Patents  
For patents covering National Instruments products, refer to the appropriate location: Help»Patents in your software, the patents.txtfile  
on your CD, or ni.com/patents.  
WARNING REGARDING USE OF NATIONAL INSTRUMENTS PRODUCTS  
(1) NATIONAL INSTRUMENTS PRODUCTS ARE NOT DESIGNED WITH COMPONENTS AND TESTING FOR A LEVEL OF  
RELIABILITY SUITABLE FOR USE IN OR IN CONNECTION WITH SURGICAL IMPLANTS OR AS CRITICAL COMPONENTS IN  
ANY LIFE SUPPORT SYSTEMS WHOSE FAILURE TO PERFORM CAN REASONABLY BE EXPECTED TO CAUSE SIGNIFICANT  
INJURY TO A HUMAN.  
(2) IN ANY APPLICATION, INCLUDING THE ABOVE, RELIABILITY OF OPERATION OF THE SOFTWARE PRODUCTS CAN BE  
IMPAIRED BY ADVERSE FACTORS, INCLUDING BUT NOT LIMITED TO FLUCTUATIONS IN ELECTRICAL POWER SUPPLY,  
COMPUTER HARDWARE MALFUNCTIONS, COMPUTER OPERATING SYSTEM SOFTWARE FITNESS, FITNESS OF COMPILERS  
AND DEVELOPMENT SOFTWARE USED TO DEVELOP AN APPLICATION, INSTALLATION ERRORS, SOFTWARE AND HARDWARE  
COMPATIBILITY PROBLEMS, MALFUNCTIONS OR FAILURES OF ELECTRONIC MONITORING OR CONTROL DEVICES,  
TRANSIENT FAILURES OF ELECTRONIC SYSTEMS (HARDWARE AND/OR SOFTWARE), UNANTICIPATED USES OR MISUSES, OR  
ERRORS ON THE PART OF THE USER OR APPLICATIONS DESIGNER (ADVERSE FACTORS SUCH AS THESE ARE HEREAFTER  
COLLECTIVELY TERMED “SYSTEM FAILURES”). ANY APPLICATION WHERE A SYSTEM FAILURE WOULD CREATE A RISK OF  
HARM TO PROPERTY OR PERSONS (INCLUDING THE RISK OF BODILY INJURY AND DEATH) SHOULD NOT BE RELIANT SOLELY  
UPON ONE FORM OF ELECTRONIC SYSTEM DUE TO THE RISK OF SYSTEM FAILURE. TO AVOID DAMAGE, INJURY, OR DEATH,  
THE USER OR APPLICATION DESIGNER MUST TAKE REASONABLY PRUDENT STEPS TO PROTECT AGAINST SYSTEM FAILURES,  
INCLUDING BUT NOT LIMITED TO BACK-UP OR SHUT DOWN MECHANISMS. BECAUSE EACH END-USER SYSTEM IS  
CUSTOMIZED AND DIFFERS FROM NATIONAL INSTRUMENTS' TESTING PLATFORMS AND BECAUSE A USER OR APPLICATION  
DESIGNER MAY USE NATIONAL INSTRUMENTS PRODUCTS IN COMBINATION WITH OTHER PRODUCTS IN A MANNER NOT  
EVALUATED OR CONTEMPLATED BY NATIONAL INSTRUMENTS, THE USER OR APPLICATION DESIGNER IS ULTIMATELY  
RESPONSIBLE FOR VERIFYING AND VALIDATING THE SUITABILITY OF NATIONAL INSTRUMENTS PRODUCTS WHENEVER  
NATIONAL INSTRUMENTS PRODUCTS ARE INCORPORATED IN A SYSTEM OR APPLICATION, INCLUDING, WITHOUT  
LIMITATION, THE APPROPRIATE DESIGN, PROCESS AND SAFETY LEVEL OF SUCH SYSTEM OR APPLICATION.  
Conventions  
The following conventions are used in this manual:  
[ ]  
Square brackets enclose optional items—for example, [response].  
»
The » symbol leads you through nested menu items and dialog box options  
to a final action. The sequence File»Page Setup»Options directs you to  
pull down the File menu, select the Page Setup item, and select Options  
from the last dialog box.  
This icon denotes a note, which alerts you to important information.  
bold  
Bold text denotes items that you must select or click in the software, such  
as menu items and dialog box options. Bold text also denotes parameter  
names.  
italic  
Italic text denotes variables, emphasis, a cross-reference, or an introduction  
to a key concept. Italic text also denotes text that is a placeholder for a word  
or value that you must supply.  
monospace  
Text in this font denotes text or characters that you should enter from the  
keyboard, sections of code, programming examples, and syntax examples.  
This font is also used for the proper names of disk drives, paths, directories,  
programs, subprograms, subroutines, device names, functions, operations,  
variables, filenames, and extensions.  
monospace bold  
Bold text in this font denotes the messages and responses that the computer  
automatically prints to the screen. This font also emphasizes lines of code  
that are different from the other examples.  
 
Chapter 1  
Document Organization...................................................................................1-1  
Bibliographic References ................................................................................1-2  
Related Publications........................................................................................1-2  
Chapter 2  
pfscale( )..........................................................................................................2-16  
optscale( ) ........................................................................................................2-17  
Reducibility....................................................................................................................2-17  
Worst-Case Performance Degradation (wcgain) ...........................................................2-18  
Conversion to a Stability Margin Problem......................................................2-18  
wcgain( )..........................................................................................................2-19  
© National Instruments Corporation  
v
MATRIXx Xmath Robust Control Module  
Contents  
Chapter 3  
Singular Value Bode Plots............................................................................................. 3-1  
L Infinity Norm (linfnorm)............................................................................................ 3-3  
Singular Value Bode Plots of Subsystems .................................................................... 3-7  
Chapter 4  
fsregu( )........................................................................................................... 4-14  
fsesti( )............................................................................................................. 4-16  
Frequency-Shaped Control Design Commands.............................................. 4-17  
Loop Transfer Recovery (lqgltr) ................................................................................... 4-22  
lqgltr( ) ............................................................................................................ 4-23  
Appendix A  
Appendix B  
Technical Support and Professional Services  
Index  
MATRIXx Xmath Robust Control Module  
vi  
ni.com  
1
Introduction  
The Xmath Robust Control Module (RCM) provides a collection of  
analysis and synthesis tools that assist in the design of robust control  
systems.  
This chapter starts with an outline of the manual and some use notes. It  
continues with an overview of the Xmath Robust Control Module (RCM)  
functions.  
Using This Manual  
This manual provides complete documentation for all the RCM functions  
examples.  
Document Organization  
This manual includes the following chapters:  
Chapter 1, Introduction, describes the Robust Control Module (RCM)  
and shows the RCM function structure.  
tools and introduces the concepts of uncertainty, robustness, and  
performance degradation in the framework of closed-loop systems.  
interested in robustness analysis or performance degradation, which  
are explained in the Stability Margin (smargin) section and the  
Worst-Case Performance Degradation (wcbode) section. The  
Advanced Topics section provides additional information but this  
material is not prerequisite to the use of RCM functions.  
Chapter 3, System Evaluation, describes system analysis functions that  
create singular value Bode plots, performance plots, and calculate the  
Lnorm of a linear system. This chapter should be of interest to all  
users.  
Chapter 4, Controller Synthesis, discusses synthesis tools in two  
categories, Hand H2. This manual does not attempt to explain all  
of the theory of H, LQG/LTR, and frequency shaped LQG design  
© National Instruments Corporation  
1-1  
MATRIXx Xmath Robust Control Module  
       
Chapter 1  
Introduction  
techniques. The general problem setup is explained together with  
known limitations; the rest is left to the references.  
Bibliographic References  
Throughout this document, bibliographic references are cited with  
bracketed entries. For example, a reference to [DoS81] corresponds  
to a document published by Doyle and Stein in 1981. For a table of  
bibliographic references, refer to Appendix A, Bibliography.  
Commonly-Used Nomenclature  
This manual uses the following general nomenclature:  
Matrix variables are generally denoted with capital letters; vectors are  
represented in lowercase.  
G(s) is used to denote a transfer function of a system where s is the  
Laplace variable. G(q) is used when both continuous and discrete  
systems are allowed.  
H(s) is used to denote the frequency response, over some range of  
frequencies of a system where s is the Laplace variable. H(q) is used to  
indicate that the system can be continuous or discrete.  
A single apostrophe following a matrix variable, for example, x',  
denotes the transpose of that variable. An asterisk following a matrix  
variable (for example, A*) indicates the complex conjugate, or  
Hermitian, transpose of that variable.  
Related Publications  
For a complete list of MATRIXx publications, refer to Chapter 2,  
MATRIXx Publications, Help, and Online Support, of the MATRIXx  
Getting Started Guide. The following documents are particularly useful  
for topics covered in this manual:  
MATRIXx Getting Started Guide  
Xmath User Guide  
Xmath Control Design Module  
Xmath Interactive Control Design Module  
Xmath Interactive System Identification Module, Part 1  
Xmath Interactive System Identification Module, Part 2  
Xmath Model Reduction Module  
MATRIXx Xmath Robust Control Module  
1-2  
ni.com  
       
Chapter 1  
Introduction  
Xmath Optimization Module  
Xmath Robust Control Module  
Xmath Xμ Module  
MATRIXx Help  
Robust Control Module function reference information is available in the  
MATRIXx Help. The MATRIXx Help includes all Robust Control functions.  
Each topic explains a function’s inputs, outputs, and keywords in detail.  
Refer to Chapter 2, MATRIXx Publications, Help, and Online Support, of  
the MATRIXx Getting Started Guide for complete instructions on using the  
Help feature.  
Overview  
RCM functionality is structured as shown in Figure 1-1.  
© National Instruments Corporation  
1-3  
MATRIXx Xmath Robust Control Module  
     
Chapter 1  
Introduction  
Analysis Functions  
smargin  
wcbode  
wcgain  
ssv  
pfscale  
optscale  
osscale  
lqgltr  
Synthesis Functions  
fslqgcomp  
fsesti  
fsregu  
hinfcontr  
singriccati  
clsys  
Utility Functions  
linfnorm  
perfplots  
Figure 1-1. RCM Function Structure  
Many RCM functions are based on state-of-the-art algorithms implemented  
in cooperation with researchers at Stanford University. The robustness  
analysis functions are based on structured singular value calculations.  
The synthesis tools expand on existing LQG (H2) techniques (LQG/LTR  
and frequency shaping) while adding new Hdesign functions.  
MATRIXx Xmath Robust Control Module  
1-4  
ni.com  
 
2
Robustness Analysis  
This chapter describes RCM tools used for analyzing the robustness  
of a closed-loop system. The chapter assumes that a controller has been  
designed for a nominal plant and that the closed-loop performance of  
this nominal system is acceptable. The goal of robustness analysis is to  
determine whether the performance will remain acceptable if the plant  
differs from the nominal plant.  
Modeling Uncertain Systems  
This section describes the method RCM uses to model an uncertain system.  
The closed-loop system is modeled as a known or nominal closed-loop  
system with input w and output z, together with k unknown or uncertain  
transfer functions δ1(jω), , δk(jω), as shown in Figure 2-1.  
Uncertain Transfer Function  
Known Nominal  
System  
q1  
r1  
δ1  
w
z
q2  
r2  
δ2  
Figure 2-1. Model of an Uncertain System  
The following transfer functions are assumed to be stable:  
δi( jω) ≤ li( jω)  
(2-1)  
where the li are given non-negative functions of frequency. This type of  
uncertainty model is known as structured nonparametric uncertainties.  
To describe this model, you also must describe the nominal closed-loop  
© National Instruments Corporation  
2-1  
MATRIXx Xmath Robust Control Module  
                   
Chapter 2  
Robustness Analysis  
system, including how the uncertain transfer functions are connected to the  
system and the magnitude bound functions li(w).  
To do this, extract the uncertain transfer functions and collect them into a  
k-input, k-output transfer matrix Δ, where:  
Δ( jω) = diagonal1( jω),...,δk( jω))  
(2-2)  
The resulting closed-loop system can be viewed as a feedback connection  
of the nominal closed-loop system with transfer matrix H(jω) and the  
uncertain transfer matrix Δ(jω). You describe your nominal closed-loop  
system as a linear system with  
w
r
z
input  
and output  
.
q
Note The signals r and q are not really inputs and outputs of the nominal system; r and q  
show how the uncertain transfer functions connect to your nominal system. The signals r  
and q each have k components.  
You will partition H into the four submatrices,  
Hzw Hzr  
H =  
Hqw Hqr  
so that Hzw is the nominal transfer matrix from w to z, Hzr is the nominal  
transfer matrix from r to z, Hqw is the nominal transfer matrix from w to q,  
and Hqr is the nominal transfer matrix from r to q.  
The magnitude bound functions li(jω) from Equation 2-1 are described  
with the PDM delb:  
ω1 l11) … lk1)  
DELB =  
,
:
:
:
ωm l1m) … lkm)  
Thus, a complete description of your system requires the system SysH  
to represent Hjw and the response delbto represent the bounds.  
MATRIXx Xmath Robust Control Module  
2-2  
ni.com  
       
Chapter 2  
Robustness Analysis  
Stability Margin (smargin)  
Assume that the nominal closed-loop system is stable. That belief raises a  
question: Does the system remain stable for all possible uncertain transfer  
functions that satisfy the magnitude bounds (Equation 2-1)? If so, the  
system is said to be robustly stable. If the magnitude bounds are small  
enough, the uncertainties will not destabilize the system; your system will  
be robustly stable.  
Roughly speaking, the stability margin of your system is defined as the  
factor by which you can increase all the magnitude bounds li and still  
maintain stability for all possible uncertain transfer functions δi. If this  
number is larger than one (0 dB), then you know that there are no uncertain  
transfer functions that satisfy the magnitude bound and destabilize your  
system. Moreover, the number tells you how much more uncertainty your  
system could tolerate than the given bounds li(ω). If the margin is less than  
one, then there are uncertain transfer functions that satisfy the magnitude  
bound (Equation 2-1) and result in an unstable system. In this case, the  
margin tells you how much you must reduce the magnitude bounds before  
you have robust stability.  
More precisely, the stability margin at frequency ω is defined as the  
smallest α such that the system can have a pole at jω, with the uncertain  
transfer functions satisfying i(jω)| ≤ αli(ω):  
margin(w) = min{ α| systems can have a pole at jω with magnitude bounds αli(jω) }  
The stability margin also can be expressed as:  
margin(w) = min{ α| det I HqrjωΔ ≠ 0 such that ii| ≤ αli(α) }  
Note The stability margin only depends on Hqr.  
The margin often is expressed in dB. If the margin is greater than zero for  
all frequencies, then your system is robustly stable. If the margin is less  
than zero for some frequencies, then your system is not robustly stable.  
In particular, there are uncertain transfer functions that satisfy the  
magnitude bound (Equation 2-1) and cause the system to have a pole at  
those frequencies where the margin is negative. This does not mean that any  
δi values that satisfy the magnitude bound will destabilize the system: it  
means that there are some bad δi values that satisfy the magnitude bounds  
and destabilize the system.  
© National Instruments Corporation  
2-3  
MATRIXx Xmath Robust Control Module  
         
Chapter 2  
Robustness Analysis  
smargin( )  
marg = smargin(SysH, delb {scaling, graph})  
The smargin( )function plots an approximation to the stability margin  
of the system as a function of frequency. For a full discussion of  
smargin( )syntax, refer to the MATRIXx Help. The approximation is  
exact if the number of uncertain transfer functions is less than four and  
scaling="OPT"(optimum scaling).  
In other cases, the approximation is generally considered to be extremely  
good. Refer to the Approximation with Scaled Singular Values section. The  
approximation is always conservative. smargin( )always will report a  
margin that is less than or equal to the actual margin.  
The smargin( )function counts the columns in delbto calculate the  
number of uncertainties k. It then assumes that the last k inputs of SysHare  
signal r in Figure 2-2, and the last k outputs are signal q. To create a Nominal  
System, refer to the Creating a Nominal System section.  
w
r
z
Known Closed-Loop System  
H(s)  
q
size k  
size k  
Figure 2-2. Nominal Closed-Loop System  
Creating a Nominal System  
To better understand how to create H(s) in Figure 2-3, you will examine  
a SISO tracking system with three uncertainties. δ1 is a multiplicative  
actuator uncertainty, while δ2 and δ3 are multiplicative sensor uncertainties.  
MATRIXx Xmath Robust Control Module  
2-4  
ni.com  
         
Chapter 2  
Robustness Analysis  
reference  
+
error  
1
x1  
x2  
+
+
1
s
1
s
8
+
reference  
2
1
K1 = 4  
K2 = 8  
Figure 2-3. SISO Tracking System with Three Uncertainties  
The H system will have the reference input as input1 and the error output  
as output1 (w and z, respectively, in Figure 2-2). Removing the δ values will  
create inputs 2 through 4 and outputs 2 through 4 (r and q, respectively, in  
Figure 2-2).  
1. The A, B, C, D matrices of the state-space system representing H are  
as follows:  
A=[-4,-8;1,0];  
B=[8,1,-4,-8;zeros(1,4)];  
C=[0,-1;-4,-8;1,0;0,1];  
D=[1,0,0,0;8,0,-4,-8;zeros(2,4)];  
H = system(A,B,C,D,{inputNames=["reference",  
"r1","r2","r3"],outputNames=["error",  
"q1","q2","q3"],stateNames=["x1","x2"]});  
2. Specify the uncertainty bounds.  
The sensor uncertainty δ3 is known to be bounded by l3(w), according  
to Equation 2-1. Because the position x2 sensor model is known to be  
accurate to 10% up to one radian per second, and very inaccurate at  
high frequencies, the l3 shown in Figure 2-4 is selected.  
© National Instruments Corporation  
2-5  
MATRIXx Xmath Robust Control Module  
 
Chapter 2  
Robustness Analysis  
10  
0
–20  
0.1  
30  
Frequency, Radian/Second  
100  
1
Figure 2-4. Bound for Sensor Uncertainty  
Note A value of l3 at one radian per second of –20 dB indicates that modeling  
uncertainties of up to 10% (–20 dB = 0.1) are allowed.  
The actuator and sensor uncertainties δ1 and δ2 are bounded by –20 dB  
at all frequencies. You will use these values to interpolate to obtain l3.  
First, create the bound for δ3 in Hz.  
L3 = pdm([-20,-20,10,10],[0.1,1,30,100]/2/pi);  
3. Now interpolate to obtain 30 points:  
L3 = interpolate(L3,logspace(0.01,10,30),{xlog});  
4. Create L1 and L2 (bounds for δ1 and δ2 ):  
L1=-20*ones(L3); L2 = L1;  
delb = [L1,L2,L3];  
5. Calculate the stability margin:  
marg=smargin(H,delb);  
smargin --> Scaling algorithm is type: PF  
smargin --> Margin computation 10% complete  
smargin --> Margin computation 50% complete  
smargin --> Margin computation 90% complete  
The output indicates that Perron-Frobenius scaling (the default) is  
used. Refer to the Approximation with Scaled Singular Values section.  
The stability margin plot is shown in Figure 2-5. The minimum margin  
is about 8 dB at about 1/2 Hz. This implies that all three l1 values  
(uncertainty bounds) could be increased (relaxed) simultaneously  
by 8 dB, and the system would still remain robustly stable.  
MATRIXx Xmath Robust Control Module  
2-6  
ni.com  
 
Chapter 2  
Robustness Analysis  
Figure 2-5. Stability Margin  
Now examine the effect on the stability margin of discretizing H(s) at  
100 Hz.  
dt = 0.01;  
Hd = discretize(H,dt);  
margD = smargin(Hd,delb);  
smargin --> Scaling algorithm is type: PF  
smargin --> Margin computation 10% complete  
smargin --> Margin computation 50% complete  
smargin --> Margin computation 90% complete  
100 Hz is a high discretization frequency for H, so the stability margin  
is unchanged in the discrete-time case. The new plot is not much  
different from Figure 2-6. Again, minimum margin is about 8 dB  
at about 1/2 Hz.  
© National Instruments Corporation  
2-7  
MATRIXx Xmath Robust Control Module  
 
Chapter 2  
Robustness Analysis  
Worst-Case Performance Degradation (wcbode)  
Even if a system is robustly stable, the uncertain transfer functions still can  
have a great effect on performance. Consider the transfer function from the  
qth input, wq, to the pth output, zp. With δ1 = ... = ...δk = 0, you have the  
nominal system, and this transfer function is the p,q entry of Hzw. This is  
called the nominal transfer function.  
When the δ values are not zero, the transfer function from wq to zp is the p,q  
entry of Hpert given by the formula:  
Hpert = Hzw + HzrΔ(I HqrΔ)1Hqw  
This is referred to as the perturbed transfer function. The perturbed transfer  
function depends on the particular δ1, …, δk.  
If the magnitude bounds are small enough, then you expect the perturbed  
transfer function Hpert to be close to the nominal transfer function. Roughly  
speaking, small perturbations should not significantly alter the closed-loop  
transfer function from wq to zp.  
The worst-case gain is defined as the largest magnitude of the perturbed  
transfer function, considering all δ values that satisfy the magnitude bound.  
More precisely:  
wcgain(ω) = max{ Hpert,pq  
Δ = diagonal1,...,δk), δi li(ω)} (2-3)  
wcgain(ω) is always larger than the nominal gain, |Hzw,pq(jω)|. This is not  
because the uncertain transfer functions only can increase the magnitude of  
the transfer function from wq to zq. In fact, it is possible that for a lucky  
choice of the δ values, the perturbed transfer function actually can be  
smaller than the nominal transfer function over all frequencies. But in the  
worst-case gain, you consider only the worst possible δ values, and these  
always increase the perturbed gain over the nominal gain.  
Intuitively, if the stability margin is large, then the uncertain transfer  
functions should not greatly effect the gain from wq to zp, so that wcgain(ω)  
should be not much larger than the nominal gain |Hzw,pq(jω)|. If the stability  
margin is small, however, wcgain(ω) could be much larger than the nominal  
gain. An extreme case occurs if the stability margin is negative (in dB) at  
the frequency δ. Then you have wcgain(ω) = , although wcbode( )clips  
the worst-case gain curve so that it never exceeds (the maximum nominal  
gain) * 100, or +20 dB. Of course, instability is an extreme form of  
performance degradation.  
MATRIXx Xmath Robust Control Module  
2-8  
ni.com  
               
Chapter 2  
Robustness Analysis  
wcbode( )  
[WCMAG, NOMMAG] = wcbode (SysH, delb, {input, output,  
graph})  
The wcbode( )function computes and plots the worst-case gain of a  
closed-loop transfer function.  
This function is useful for checking a system that already has been verified  
to be robustly stable using smargin( ). For example, a system can have a  
minimum stability margin of 4 dB, so it is robustly stable. If the worst-case  
gain from a function input to the output it commands has a 20 dB peak, then  
even though the system is robustly stable, the design is unacceptable. On  
the other hand, if you verify that the perturbed closed-loop transfer function  
increases only 2 dB over the nominal, then the design is probably  
acceptable.  
The wcbode( )function computes and plots an approximation to  
wcgain(ω), the largest possible magnitude of a perturbed closed-loop  
transfer function that can be caused by uncertain transfer functions that  
satisfy the magnitude bound. The wcbode( )function is conservative:  
it does not under-report the maximum of the perturbed transfer function.  
A large value of wcbode( )indicates instability: wcgain(ω) = . In this  
case, wcbode( )returns a maximum value of ten times the maximum of  
the nominal transfer function over all frequencies. Consequently, the  
window is clipped at 20 dB above the maximum of the nominal transfer  
function over all frequencies. The wcbode( )function also plots the  
nominal transfer function for reference.  
Using wcbode( ) to Analyze Performance Degradation  
The wcbode( )function can be used to analyze performance degradation  
for the system you have been using (Figure 2-3). The transfer function,  
which should be small, is from reference to error (input 1 to output 1).  
Figure 2-6 shows the results of the following function call:  
[NOMMAG,WCMAG]=wcbode(H,delb,{input=1,output=1});  
The performance degradation due to the uncertainties is small but not  
negligible.  
© National Instruments Corporation  
2-9  
MATRIXx Xmath Robust Control Module  
   
Chapter 2  
Robustness Analysis  
Figure 2-6. Performance Degradation of the SISO Tracking System  
Advanced Topics  
This section describes the theoretical background on robustness analysis  
and performance degradation.  
Stability Margin  
This section discusses advanced aspects of computing the stability margin  
and the related scaling algorithms.  
Stability Margin and Structured Singular Values (μ)  
The stability margin was first defined by Safonov in [Saf82]. If you let  
M = Hqrdiagonal(l1(w), ...,lk(w))  
then you can express the margin at frequency d as  
margin(ω) = max  
det(I MΔ) ≠ 0  
MATRIXx Xmath Robust Control Module  
2-10  
ni.com  
           
Chapter 2  
Robustness Analysis  
for all diagonal Δ such that  
1
( Δii ≤ α)} = -------------  
μ(M)  
where μ(.) is the structured singular value, introduced by Doyle in  
[Doy82]. Thus, the margin is the inverse of the structured singular value of  
Hqr diagonally scaled by the magnitude bounds.  
There is no numerically efficient algorithm that is guaranteed to compute  
μ(M), and hence the stability margin. However, it is possible to compute  
various good approximations to μ(M). One of these approximations is often  
exact.  
Stability Margin Bounds Using Singular Values  
A popular but conservative method uses singular values:  
1
----------------------  
margin(ω) ≥  
(2-4)  
σ
(M)  
max  
Plotting the right side of Equation 2-4 gives a lower bound on the  
actual stability margin. To get this plot, specify smargin( )with  
scaling="SVD". This approximation can be very conservative, meaning  
that the left side can be much larger than the right side. This fact spurred  
the study of structured singular values and the other approximations  
discussed in the following sections.  
Use of Scaling Example  
For this example, you will use the system in Figure 2-3. This time  
smargin( )will be invoked with scaling="SVD", so smargin( )  
will calculate Equation 2-4.  
margSVD = smargin(H,delb,{scaling="SVD"});  
smargin --> Scaling algorithm is type: SVD  
smargin --> Margin computation 10% complete  
smargin --> Margin computation 50% complete  
smargin --> Margin computation 90% complete  
© National Instruments Corporation  
2-11  
MATRIXx Xmath Robust Control Module  
         
Chapter 2  
Robustness Analysis  
You can compare this margin to that of the example in the Creating a  
Nominal System section; the following inputs produce Figure 2-7.  
plot ([marg,margSVD],{xlog}  
legend=["PF_SCALE","SVD"],  
ylab="Stability Margin,dB",  
xlab="Frequency, Hz."})  
Figure 2-7. pfscale( ) versus svd Stability Margins  
Note The singular value approach gives results that are too conservative, suggesting that  
the uncertainties can destabilize the system. Conversely, you know from the scaled singular  
value calculations that the system is robustly stable.  
Approximation with Scaled Singular Values  
In [Saf82] and [Doy82], the inequality  
min  
σmax(DMD1) ≥ μ(M)  
(2-5)  
diagonal  
is noted. This optimization problem can be shown to be  
unimodal—for D > 0, an assumption that can be made without loss  
MATRIXx Xmath Robust Control Module  
2-12  
ni.com  
     
Chapter 2  
Robustness Analysis  
of generality—so, roughly speaking, it can be solved. [SD83,SD84]  
discusses this optimization problem.  
Notice that:  
σmax(M) = σ(DMD1)  
for  
D = 1  
so you have the following from Equation 2-5:  
σmax(M) ≥ μ(M)  
This inequality is thought to be nearly an equality, so that the left side is a  
good engineering approximation to the right side. No theory supports this  
generally held belief, but no example is known where the left side is more  
than 15% larger than the right side. Equality can be shown to hold provided  
k 3—for example, if there are three or fewer uncertain transfer functions  
[Doy82].  
Note The approximation equation of μ(M) (Equation 2-5) is an upper bound. This means  
that the stability margin calculated using this approximation is conservative, that is, less  
than the actual stability margin. This optimization problem itself can be difficult. Osborne  
[Osb60] and Safonov [Saf82] provide two methods for finding good suboptimal scalings  
for Equation 2-5.  
Both Osborne’s and Safonov’s Perron-Frobenius scalings usually have  
been found to be close to the optimum for the optimization problem  
equation. The resulting approximations,  
1  
ˆ
uOS(M) = σmax(DOSMDOS  
)
)
1  
ˆ
uPF(M) = σmax(DPFMDPF  
are thought to be good engineering approximations to μ. optscale( )  
provides an iterative optimization function based on the ellipsoid  
algorithm.  
© National Instruments Corporation  
2-13  
MATRIXx Xmath Robust Control Module  
Chapter 2  
Robustness Analysis  
Comparing Scaling Algorithms  
Using the system from the first example (Figure 2-3), you can compare  
the results of using the three scaling algorithms:  
MARG_PF=smargin(H,delb,{scaling="PF",!graph});  
MARG_OS=smargin(H,delb,{scaling="OS",!graph});  
MARG_OPT=smargin(H,delb,{scaling="OPT",!graph});  
plot ([MARG_PF,MARG_OS,MARG_OPT],{xlog,  
legend=["PF","OS","OPT"],xlab="Frequency, Hz.",  
ylab="Stability margin, dB"})  
Figure 2-8 plots the margins produced by the three scaling algorithms.  
Notice that in this problem the three scalings yield identical stability  
margins.  
Figure 2-8. Results of Scaling Algorithm Options  
MATRIXx Xmath Robust Control Module  
2-14  
ni.com  
 
Chapter 2  
Robustness Analysis  
ssv( )  
[v,vD] = SSV(M, {scaling})  
The ssv( )function computes an approximation (and guaranteed upper  
bound) to the Scaled Singular Value of a complex square matrix M, where  
M can be a reducible matrix. The scaled singular value v(M) is defined by:  
v(M) =  
inf  
σ(DMD1)  
n × n  
D C  
, det(D) ≠ 0, diagonal  
Scaling can be accomplished with one of three algorithms:  
Perron-Frobenius—If {scaling="PF"}Safonov’s  
Perron-Frobenius method [Saf82] is used. This method finds the scaled  
singular value for non-negative real matrices M. In general, it is  
suboptimal if M is complex. This algorithm is the default because  
empirical tests show that is the fastest of the three.  
Osborne—If {scaling="OS"}, Osborne’s Method [Osb60] is used.  
This method solves the problem of finding DO such that  
inf  
diagonal  
DOMDO1  
DMD1  
F
D
where D is diagonal and positive, and is the Frobenius norm.  
F
Thus, the Osborne method minimizes the Frobenius norm, and is  
therefore suboptimal.  
Optimal—If {scaling="OPT"}, Boyd’s ellipsoid algorithm  
[BYB89] is used. This algorithm computes the scaled singular value  
to a guaranteed accuracy. It is, however, the most computationally  
expensive of the three algorithms.  
ssv( ) Examples  
Consider the complex matrix M:  
M = [–1, jay, 0; 0, 2*jay, 1+jay;1, 0, 1];  
ssv( )can return the optimally scaled singular value of M using Osborne,  
Perron-Frobenius, or Boyd methods:  
VOS=ssv(M,{scaling="OS"})  
VOS (a scalar) = 2.56723  
VPF=ssv(M,{scaling="PF"})  
VPF (a scalar) = 2.45133  
© National Instruments Corporation  
2-15  
MATRIXx Xmath Robust Control Module  
           
Chapter 2  
Robustness Analysis  
VOPT=ssv(M,{scaling="OPT"})  
VOPT (a scalar) = 2.43952  
VSVD = max(svd(M))  
VSVD (a scalar) = 2.65886  
osscale( )  
[v, vD] = osscale(M)  
The osscale( )function scales a matrix using the Osborne Algorithm.  
A diagonal scaling DOS is found that minimizes the Frobenius norm of  
DOSMDO1S , which is the square root of the sum of the squares of its  
singular values. If M is reducible, osscale( )may encounter a divide  
by zero. To avoid this, use ssv( )with the Osborne scaling option:  
[v,vD]=ssv(M,{scaling="OS"})  
pfscale( )  
[v, vD] = pfscale(M)  
The pfscale( )function scales a matrix using the Perron-Frobenius  
Algorithm. This scaling is optimal for matrices with all positive entries.  
The matrix M must be irreducible for this function. If M is reducible,  
use ssv( )with the Perron-Frobenius scaling option instead:  
[v,vD]=ssv(M,{scaling="PF"})  
The optimum diagonal scaling is found for M using the Perron-Frobenius  
theory of non-negative matrices. This scaling is given by  
pi  
DPi F  
=
---  
qi  
where p and q are right and left eigenvectors of | associated with its largest  
eigenvalue:  
Mp = λmaxp,  
MTq = λmaxq,  
p 0 q  
MATRIXx Xmath Robust Control Module  
2-16  
ni.com  
       
Chapter 2  
Robustness Analysis  
optscale( )  
[v, vOPTD] = optscale (M, {tol})  
The optscale( )function optimally scales a matrix. An iterative  
optimization (ellipsoid) algorithm which calculates upper and lower  
bounds on the left side of Equation 2-5 is used. If these bounds are within  
a relative accuracy you have specified (tol), optscale( )stops.  
optscale( )also will stop and issue a warning if the maximum number  
of iterations is reached:  
200 × rows(M)  
optscale( )will find a μ(M) no larger than pfscale( ).  
Reducibility  
In some cases, the uncertain transfer functions can be divided into groups  
that do not interact. This is illustrated in Figure 2-9.  
δ1  
δ2  
δ3  
δ4  
Figure 2-9. Non-Interacting Uncertain Transfer Functions  
As you can see, δ3 does not affect the stability margin at all because there  
is no feedback through it. The system in Figure 2-9 can be reduced to the  
two separate systems shown in Figure 2-10. The stability margin of  
Figure 2-9 is the minimum of the stability margins of the systems in  
Figure 2-10.  
© National Instruments Corporation  
2-17  
MATRIXx Xmath Robust Control Module  
         
Chapter 2  
Robustness Analysis  
δ1  
δ4  
δ2  
Figure 2-10. Reduction to Separate Systems  
In terms of the approximations to the margin discussed above, this  
reducibility will manifest itself as a problem such as divide-by-zero or  
nontermination. It really means that the minimum of the optimization  
problem is not achieved by any finite scaling.  
A matrix M can be split into its reducible components using the following  
technique (refer to[BeP79]):  
1. Form the matrix X = (αI + M) – 1 for any α larger than the spectral  
radius of M, for example 2 M .  
2. Form Y = X + XT where Y has a positive i,j entry if and only if δi and δj  
are in the same reduced system; otherwise, the entries will be zero.  
ssv( )checks for reducibility before invoking a scaling algorithm. The  
margins of each of the reduced systems then can be calculated separately,  
and the minimum taken.  
Conversion to a Stability Margin Problem  
In [DWS82], it is shown that a simple relation holds between the  
worst-case gain defined in Equation 2-3, and the stability margin. For γ > 0,  
wcgain (jw) £ g if and only if m(Hred(jw) diag(g-1,  
l1(w),...,lk(w)) £ 1  
where Hred is H with the rows and columns corresponding to all inputs and  
outputs deleted except the ones of interest (the qth input and the pth output).  
This can be interpreted as adding a fictitious uncertain transfer function  
from wq to zp with magnitude bound γ–1 at the given frequency. This  
additional uncertainty is called a performance loop as described in  
reference [BoB91].  
MATRIXx Xmath Robust Control Module  
2-18  
ni.com  
       
Chapter 2  
Robustness Analysis  
Using this relation and any of the previously discussed approximations for  
μ(.), you can compute an approximation to wcgain( ). Because the  
approximations to μ(.) are upper bounds, the resulting approximations to  
wcgain( )also are upper bounds. For speed purposes, wcgain( )uses  
Perron-Frobenius scaling to calculate the approximation of μ.  
wcgain( )  
gamma = wcgain(H, gammin, gammax, gam0)  
The wcgain( )function estimates the largest possible magnitude from  
a given input of the system to a given output, when the other inputs are  
connected to the other outputs through uncertain transfer functions  
bounded by one.  
For a discussion of wcgain( )syntax, refer to the Xmath Help. This is a  
low-level function that calculates the worst-case gain at a single frequency,  
where the magnitude bounds are normalized to one as follows:  
l1 = = lk = 1  
Because it is a lower level function, there is no syntax checking. This  
function is called by the wcbode( )function.  
© National Instruments Corporation  
2-19  
MATRIXx Xmath Robust Control Module  
   
3
System Evaluation  
This chapter describes system analysis functions that create singular value  
Bode plots, performance plots, and calculate the Lnorm of a linear  
system.  
Singular Value Bode Plots  
The singular value Bode plot is a MIMO generalization of the bode( )  
magnitude plot. It is calculated as  
σi(H(jw)),  
i = 1k  
where  
and  
k = min((ninputs,noutputs))  
σ1(H) ≥ σ2(H) ≥ … ≥ σk(H) ≥ 0  
In these equations, σi(H(jw)) can be thought of as the maximum gain of the  
system at frequency ω, and σk(H(jw)) can be thought of as the minimum  
gain of the system at frequency ω.  
If σi(H(jw)) » σk(H(jw)), then at the frequency ω, the system gain can be  
large for some input directions and small for other input directions.  
The singular value plot allows you to generalize to the MIMO case notions  
such as “the command-to-tracking error transfer function is small” or “the  
loop gain is large.” For example:  
If the system represents the command-to-tracking error for a  
closed-loop system, then you would hope that σ1, and hence all  
σ values, are small over the bandwidth of the system. This means  
that the command-to-tracking error is small in all directions at these  
frequencies.  
The singular value plot of a certain transfer matrix gives a lower bound  
on the stability margin of the system.  
© National Instruments Corporation  
3-1  
MATRIXx Xmath Robust Control Module  
         
Chapter 3  
System Evaluation  
Refer to [BoB91] in Appendix A, Bibliography.  
Creating a Singular Value Plot  
Example 3-1  
1. Let a system H be a 2-input/2-output system:  
tf=makepoly([1,2],"s")/...  
polynomial([0,-2.334,-12],"s")  
tf (a transfer function) =  
s + 2  
--------------------  
(s + 2.334)(s + 12)s  
System is continuous  
H = [tf, 2*tf; tf*tf, tf+3];  
[outputs,inputs]=size(H)  
outputs (a scalar) = 2  
inputs (a scalar) = 2  
2. Now plot the singular values of the system between 0.01 and 100 Hz  
using svplot( ):  
svplot(H,{Fmin=0.01,Fmax=100})  
The result is shown in Figure 3-1. For a discussion of svplot( )syntax,  
refer to the Xmath Help.  
MATRIXx Xmath Robust Control Module  
3-2  
ni.com  
Chapter 3  
System Evaluation  
Figure 3-1. Singular Value Plot  
L Infinity Norm (linfnorm)  
The Lnorm of a stable transfer matrix H is defined as:  
H
= sup σ(H( jω))  
ω ∈ ℜ  
where σ is the maximum singular value and H(jω) is the transfer matrix  
under consideration.  
The Lnorm of a stable transfer matrix is the maximum of the maximum  
singular values over frequency. For example, the highest point of its  
singular values plot. Observe that the Lnorm can be calculated even if H is  
not stable.  
A simple interpretation of the Lnorm of a stable system can be given as:  
RMS(y)  
-------------------  
= max  
H
RMS(u)  
where u and y are the input and output of H, respectively. This means that  
is the root mean square (RMS) gain of the system: it is the largest  
H
© National Instruments Corporation  
3-3  
MATRIXx Xmath Robust Control Module  
     
Chapter 3  
System Evaluation  
factor by which the RMS value of a signal flowing through H can be  
increased.  
By comparison, the H2 norm is defined as:  
k
1
σi(H( jω))2dw  
------  
H
=
2
2π ∫ ∑  
i = 1  
This norm can be interpreted as the RMS value of the output when the input  
is unit intensity white noise. It can be computed in Xmath using the rms( )  
function.  
For discrete-time systems with a stable H,  
max  
H
=
σ(H(e jω))  
ω ∈ (π, π)  
where σ is the maximum singular value and H(ejω) is the transfer matrix  
under consideration.  
linfnorm( )  
[sigma, vOMEGA] = linfnorm( Sys, {tol,maxiter} )  
The linfnorm( )function computes the Lnorm of a dynamic system  
using a quadratically convergent algorithm. The linfnorm( )function  
relies on eigenvalue calculations of a Hamiltonian matrix with twice as  
many states as Sysand, consequently, may be unreliable for large systems.  
A singular value plot created with svplot( )can be used as an alternative  
in these cases. Refer to the Singular Value Bode Plots section.  
The keyword tolcontrols the required relative accuracy. The default  
is 0.01. maxiteris the maximum number of iterations. The default  
is 15.  
If the maximum norm is found at ω = , linfnorm( )returns:  
vOMEGA = Infinity  
sigma = gain at infinity.  
MATRIXx Xmath Robust Control Module  
3-4  
ni.com  
     
Chapter 3  
System Evaluation  
If Ahas an imaginary eigenvalue at jω0, linfnorm( )returns:  
vOMEGA = ω0  
SIGMA = Infinity  
where ω0 is one of the imaginary eigenvalues of A.  
Even if H is unstable, linfnorm( )returns its maximum singular  
value on the jω axis.  
For discrete-time systems linfnorm( )converts a discrete-time L∞  
norm computation problem to a continuous-time problem using a Cayley  
transformation. For example, it maps the unit circle conformally onto the  
complex right half plane using a linear fractional transformation. The  
linfnorm( )function then calls itself to solve the continuous-time  
problem, and finally converts the solution back to discrete-time.  
Example 3-2  
Example of linfnorm( )  
Sys=system([-0.2,-1;1,0],[1,0]',[0,1],0);  
[sigma,omega]=linfnorm(Sys)  
sigma (a scalar) = 5.07322  
omega (a scalar) = 0.157081  
The linfnorm( )function will return the Lnorm (sigma) of the transfer  
matrix H(jω) described by Sys, and omegais the vector of frequencies  
where it is achieved. linfnorm( )computation can be checked by  
plotting the singular values of H(jω) as a function of ω (Figure 3-2).  
sv=svplot(Sys,{fmin=.01, fmax=1.0});  
© National Instruments Corporation  
3-5  
MATRIXx Xmath Robust Control Module  
Chapter 3  
System Evaluation  
Figure 3-2. Singular Values of H(jω) as a Function of ω  
Note svis returned in dBs. Check that sigmais within 0.01 (the default value of tol) of  
10**(max(sv,{channels})/20).  
[sigma,10^(max(sv,{channels})/20)]  
ans (a row vector) = 5.07322 4.98731  
The linfnorm( )function also can be used on discrete-time systems.  
Consider a state-space system with a sample rate of 10 Hz:  
SysD=system([0.5,0.5;0.8,0.5],[0.8,0.5]',  
[0,1],0,{dt=0.1})  
[sigma,omega]=linfnorm(SysD)  
sigma (a scalar) = 5.99267  
omega (a scalar) = 0  
MATRIXx Xmath Robust Control Module  
3-6  
ni.com  
 
Chapter 3  
System Evaluation  
Singular Value Bode Plots of Subsystems  
To evaluate the performance achieved by a given controller rapidly, it is  
useful to check four basic maximum singular value plots—for example, the  
transfer matrices from process and sensor noises to the error and actuator  
signals.  
perfplots( )  
SV = perfplots ( Sys, nd, ne, { keywords } )  
The perfplots( )function plots the maximum singular value of the  
four transfer matrices of the system in the following figure.  
e
u
d
n
Sys  
In most applications, the perfplots( )function is applied to a system of  
the form shown in Figure 3-3, where Pis the plant and Kis a proposed  
controller.  
process  
noise  
error  
d
e
u
y
P
sensor  
noise  
n
actuator  
u
K
Sys  
Figure 3-3. Typical System with Plant and Controller  
© National Instruments Corporation  
3-7  
MATRIXx Xmath Robust Control Module  
         
Chapter 3  
System Evaluation  
The four transfer matrices are labeled e/d, e/n, u/d, and u/n in the final plot.  
The plots in the top row, consisting of e/d and e/n, show the regulation or  
tracking achieved by the controller. If both these quantities are small, then  
the disturbance d and the sensor noise n will not make the error signal e  
large.  
The bottom row of plots, consisting of u/d and u/n, show the actuator effort  
used by the controller. If these are both small, then the actuator effort u,  
which results from the disturbance d and the sensor noise n will be small.  
A classic trade-off in controller design boils down to a choice between  
making the top row of a perfplot( )small (good regulation/tracking)  
and making the bottom row small (low actuator effort). For example, by  
varying the design parameter ρ in the lqgltr( )regulator design process,  
the magnitude of the top two transfer matrices can be traded off against the  
magnitude of the bottom two. Increasing ρ makes the top two magnitudes  
smaller but makes the bottom two larger.  
The columns of a perfplot( )have a dual interpretation. The plots in the  
left column, e/d and u/d, show how sensitive the system is to the process  
noise or disturbance d. The plots in the right column, e/n and u/n, show how  
sensitive the system is to the sensor noise n. Again, there is a trade-off  
between making the magnitudes of the transfer matrices on the left small  
(good disturbance rejection) and making the magnitudes of the transfer  
matrices on the right small (low sensitivity to sensor noise). In the  
lqgltr( )estimator design, the parameter ρ controls the relative  
magnitude of the left and right plots. Increasing ρ makes the left two  
magnitudes smaller but makes the right two larger. Refer to Example 3-3.  
Example 3-3  
Example of perfplots( )  
Consider the simple closed-loop system shown in Figure 3-4.  
noise  
disturbance  
+
e
1
s
+
+
n
K
Figure 3-4. Closed-Loop System  
MATRIXx Xmath Robust Control Module  
3-8  
ni.com  
   
Chapter 3  
System Evaluation  
The system matrix can be calculated using the afeedback( )function for  
different values of K. Consider two cases: K = 1 and K = 5.  
P = 1/makepoly([1,0],"s")  
P (a transfer function) =  
1
--  
s
System is continuous  
K1= 1/makepoly(1,"s")  
K1 (a transfer function) =  
1
-
1
System is continuous  
K5= 5/makepoly(1,"s");  
Sys1 = afeedback(P,K1);  
Sys5 = afeedback(P,K5);  
The effect of the value of K on closed-loop performance can be investigated  
using perfplots( ).  
sv1 = perfplots(Sys1,1,1);  
Overlap plots:  
sv5 = perfplots(Sys5,1,1,{!graph});  
for i = 1:4  
plot(sv5(1,i),  
{graphnumber=i,line_style=2,keep})?  
endfor  
In Figure 3-5, you can see that over the bandwidth of 0.1 Hz, the controller  
K = 5 has better regulation (e/d is smaller for K = 5 than for K = 1, with e/n  
about the same for both cases) but uses slightly more actuator effort. Above  
the bandwidth of 0.1 Hz, the e/n and u/n show that the K = 5 controller is  
more sensitive to sensor noise. In classic terms, the K = 5 controller has a  
higher bandwidth.  
© National Instruments Corporation  
3-9  
MATRIXx Xmath Robust Control Module  
Chapter 3  
System Evaluation  
Figure 3-5. Perfplots( ) for K = 1 and K = 5  
clsys( )  
SysCL = clsys( Sys, SysC )  
The clsys( )function computes the state-space realization SysCL, of the  
closed-loop system from wto zas shown in Figure 3-6.  
z
y
w
u
Sys  
SysC  
Figure 3-6. Closed Loop System from w to z  
MATRIXx Xmath Robust Control Module  
3-10  
ni.com  
       
Chapter 3  
System Evaluation  
WhereSysC=system(Ac,Bc,Cc,Dc),Sys=system(A,B,C,D),andnz  
is the dimension of zand nwis the dimension of w:  
nw  
nw  
Bw  
nz  
nz  
Dzw Dzu  
Dyw Dyu  
B is  
C is  
D is  
Cz,Cy  
Bu  
Given the above, SysCLis calculated as shown in Figure 3-7.  
A+Bu(I DcDyu)1DcCy  
BcCy + BcDyu(I DcDyu)1DcCy Ac + BcDyu(I DcDyu)1Cc  
Bu(I DcDyu)1Cc  
ACL  
=
Bw + Bu(I DcDyu)1DcDyw  
BcDyw + BcDyu(I DcDyu)1DcDyw  
BCL  
=
Cz + Dzu(I DcDyu)1DcCy Dzu(I DcDyu)1Cc  
CCL  
=
DCL = Dzw + Dzu(I DcDyu)1DcDyw  
Figure 3-7. Calculation of the Closed Loop System (SysCL)  
The closed-loop system is assumed to be well-posed—(I DcDyu) must  
be invertible). A well-posed closed-loop system assures that if two given  
systems, Sysand SysC, are proper (only proper transfer functions can be  
represented in state space), then the resulting closed-loop system, SysCL,  
also is proper and therefore realizable in state space.  
Figure 3-8 is an example of an ill-posed feedback system, where the  
closed-loop transfer function is s+1, which cannot be represented as  
a state-space system.  
© National Instruments Corporation  
3-11  
MATRIXx Xmath Robust Control Module  
   
Chapter 3  
System Evaluation  
1
s
+ 1  
+
+
Figure 3-8. Ill-Posed Feedback System  
Example 3-4  
Example of Closed-Loop System  
a = 1;  
b = [1,0,1];  
c = b';  
d = [0,0,0;0,0,1;0,1,0];  
Sys = SYSTEM(a,b,c,d);  
SysC = SYSTEM(-40,2.7,-40,0);  
SysCL = clsys(Sys,SysC)  
SysCL (a state space system) =  
A
1
-40  
-40  
2.7  
B
1
0
0
2.7  
C
1
0
0
-40  
D
0
0
0
0
X0  
0
0
System is continuous  
MATRIXx Xmath Robust Control Module  
3-12  
ni.com  
 
4
Controller Synthesis  
This chapter discusses synthesis tools in two categories, Hand H2. This  
chapter does not explain all of the theory of H, LQG/LTR, and frequency  
shaped LQG design techniques. The general problem setup is explained  
together with known limitations.  
H-Infinity Control Synthesis  
Problem Definition  
The Hcontrol synthesis function hinfcontr( )finds a stabilizing  
multivariable controller K for the plant P, as shown in Figure 4-1.  
In the closed-loop system with plant P and controller K, all frequencies ω,  
σmax(Hew(jω)) ≤ γ  
(4-1)  
where Hew is the closed-loop transfer matrix from w to e and γ is some  
specified limit. Equation 4-1 can be expressed in terms of the Hnorm as:  
Hew ≤ γ  
Hew  
e
w
u
y
P
K
Figure 4-1. Closed-Loop System with Plant P and Controller K  
The function hinfcontr( )is based on the 2-Riccati state space  
solutions presented in [GD88,DGKF89]. You can examine these references  
for theoretical descriptions.  
© National Instruments Corporation  
4-1  
MATRIXx Xmath Robust Control Module  
                 
Chapter 4  
Controller Synthesis  
The function hinfcontr( )can be used to find an optimal Hcontroller  
K that is arbitrarily close to solving:  
min Hew ≤ γ = γopt  
(4-2)  
K
describes how the optimum can be found manually by decreasing γ until  
an error condition occurs, or conversely by increasing γ until the error  
condition is fixed.  
The particular restrictions, required by the 2-Riccati solutions and  
summarized in the Restrictions on the Extended Plant section are  
those imposed in [GD88,DGKF89].  
Extended Transfer Matrix  
Referring to Figure 4-1, plant P specifies two groupings of vector inputs  
and outputs. Such systems or transfer matrices are referred to as extended  
transfer matrices or systems. To enter these in Xmath requires a  
modification of your existing system representation. The standard system  
has the form y = G(s)u and can be described either in state-space form:  
x' = Ax + Bu  
y = Ax + Du  
or as a transfer matrix:  
G(s) = D + C(sI A)1B  
G(s) can be described in Xmath using the state-space system object:  
G = system(A,B,C,D)  
There is, however, insufficient information in this form to distinguish  
the input/output groupings in the extended system P in Figure 4-1.  
The state-space form of P is:  
·
x = Ax + B1w + B2u  
e = C1x + D11w + D12u  
y = C2x + D21w + D22u  
MATRIXx Xmath Robust Control Module  
4-2  
ni.com  
         
Chapter 4  
Controller Synthesis  
Equivalently, as a transfer matrix:  
D11 D12  
C1  
C2  
P(s) =  
+
(sI A)1[B1 B2]  
D21 D22  
To enter the extended system, you must know the sizes of e and w shown  
in Figure 4-1. The extended plant P can be constructed using the Xmath  
interconnection functions, as shown in Example 4-1.  
Building the Plant Model  
The general form of the plant P is shown in Figure 4-2.  
Plant P  
v
z
w
e
y
Win  
Wout  
G
u
Figure 4-2. Construction of Plant P  
The plant consists of three transfer matrices: Win and Wout, referred to as  
weights, and G which can be interpreted as the system dynamics. Both P  
and G distinguish inputs and outputs into two groups of variables.  
The input/output variables are organized as follows:  
actuator/sensor variables  
u—vector of actuator (control) signals  
y—vector of sensor (measured) and other accessible signals  
exogenous inputs  
v—vector of commands and disturbance  
w—vector of “normalized” commands and disturbances  
performance outputs  
z—vector of critical performance signals (regulated variables)  
e—vector of “normalized” critical performance signals  
© National Instruments Corporation  
4-3  
MATRIXx Xmath Robust Control Module  
     
Chapter 4  
Controller Synthesis  
The transfer matrix G can be viewed as a model of the underlying system  
dynamics with v and u as generalized forces that produce effects in the  
performance signals z and measured signals y.  
The weight Win is used to model the exogenous input v by v = Winw.  
Similarly, the critical performance variables in the vector z are weighted to  
form the normal critical variables e = Woutz.  
In general, the input weight Win can be viewed as a dynamic model of the  
exogenous inputs and the output weight Wout as the inverse of the desired  
performance. As an illustration, consider the plant configuration in  
Figure 4-3.  
P
G
Win  
Wout  
Wreg  
wdist  
yreg  
d
n
ereg  
w
w
u
e
e
y
Wdist  
wnoise  
eact  
u
Gdyn  
Wnoise  
Wact  
ysens  
Figure 4-3. Typical Plant Configuration  
The exogenous input vectors d and n represent disturbances and sensor  
noise, respectively. These are generated by passing normalized  
unpredictable signals, ωdist and ωnoise, through stable transfer matrices,  
Wdist and Wnoise, respectively. The critical performance variables are some  
regulated variables yreg, as well as the actuator commands u. These are  
weighed by the transfer matrices Wreg and Wact to form the normalized error  
variables ereg and eact. The sensed variables ysens are contaminated by  
additive noise n to form the measured signal y. The transfer matrix Gdyn  
represents the underlying system dynamics. Observe that the transfer  
matrix G, as defined in [BBK88], consists of Gdyn with some special  
output/input connections among the variables n and u as depicted in  
Figure 4-3. This is in the form of the familiar LQG setup, except that  
MATRIXx Xmath Robust Control Module  
4-4  
ni.com  
 
Controller Synthesis  
here the weighting matrices are transfer matrices, whereas in the LQG  
setup they are constants.  
A description of the plant in Figure 4-3 is as follows:  
Dynamical system Gdyn:  
·
x = Ax + Bdistd + Bact  
yreg = Creg  
ysens = Csens  
u
x
x
Measured variables y = ysens + n:  
Input weight Win:  
wdist  
wdist  
d
n
Wdist  
0
=
= Win  
wnoise  
wnoise  
0
Wnoise  
Output weight Wout:  
ereg  
yreg  
u
yreg  
u
Wreg  
0
=
= Win  
eact  
0
Wact  
Weight Selection  
In the standard LQG formulation, the weighting functions (Wdist, Wnoise  
,
Wreg, and Wact) are all constant matrices. In fact, if Qreg and Qact are positive  
definite matrices in the LQ cost, for example,  
1
2
--  
J =  
x'Qregx + u'Qactu dt  
0
then the following apply:  
1/2  
1/2  
Wreg = Q  
Wact = Q  
act  
reg  
Similarly, if Rproc and Rsens are positive definite matrices corresponding to  
the process and measurement noise intensities, then the following apply:  
1/2  
1/2  
Wdist = Q  
Wnoise = Q  
sens  
proc  
© National Instruments Corporation  
4-5  
MATRIXx Xmath Robust Control Module  
   
Chapter 4  
Controller Synthesis  
Selecting these weights has much the same effect here. Specifically, let Hzv  
be the closed-loop transfer matrix (with u = Kγ) from inputs:  
d
v =  
n
to outputs:  
yreg  
z =  
u
Thus,  
H
H
yreg  
yreg  
d
n
Hzv  
=
Hud Hun  
Suppose that the controller u = Ky approximates Equation 4-2. Thus,  
Wout HzvWin ≈ γopt  
In many cases, this means that the maximum singular value of the  
frequency response matrix (Wout HzvWin)( jω) is constant over all  
frequencies. That is,  
Wreg  
H
dWdist Wreg  
H
nWnoise  
yreg  
yreg  
σmax  
(jw) ≈ γopt  
WactHudWdist WactHjnWnoise  
An interpretation is that the weighting filters Win and Wout determine the  
shape of the closed-loop frequency response Hzw( jω), and γopt determines  
the peak value. This observation helps motivate the selection of the weights  
so as to shape the closed-loop frequency response matrix Hzw( jω).  
Observe, however, that the elements of the frequency response matrix,  
(WoutHzvWin)( jω), need not be constant. Instead, the maximum singular  
value of at least one of the four subblocks is within 3 dB of γopt. For all ω,  
M(ω) ≤ σmax[(Wout HzvWin)(jω)] ≤ 2M(ω)  
MATRIXx Xmath Robust Control Module  
4-6  
ni.com  
Chapter 4  
Controller Synthesis  
where  
and  
M(ω) = max{m11(u),m12(u)m21(u)m22(u)}  
m11 = σmax[WregHy dWdist  
]
reg  
m12 = σmax[WregHy nWnoise  
]
reg  
m21 = σmax[WactHudWdist  
]
m22 = σmax[WactHunWnoise  
]
The weights also can be viewed as “design knobs” (for example,  
[ONR84]). In this view, the weights are not directly related to specific  
disturbance or performance models but rather are used as a vehicle to obtain  
a closed-loop transfer matrix, Hzv, from v to z with desired properties. For  
every selection of weights Win and Wout, the closed-loop system has the  
following property:  
WinHzvWout ≤ γ  
But Hzv has other properties, both good and bad. To some extent, these all  
can be affected by varying the weights. An effective way to provide a rapid  
evaluation of performance is with the function perfplots( ), as  
described in the perfplots( ) section of Chapter 3, System Evaluation. With  
a few trial and error adjustments of the weights, perfplots( )will give  
a good indication of their effect on performance.  
Restrictions on the Extended Plant  
Not all choices of weights will result in an extended plant P = WoutGWin  
that will solve Equation 4-1. The following conditions, established in  
references [GD88,DGKF89], if satisfied, will result in a solution. If any  
are not satisfied, an error condition occurs.  
The conditions are:  
(A, B2, C2) can be stabilized and detected  
rank(D12) = NU (number of inputs)  
rank(D21) = NY (number of outputs)  
© National Instruments Corporation  
4-7  
MATRIXx Xmath Robust Control Module  
   
Chapter 4  
Controller Synthesis  
For all ω ≥ 0,  
A jωI B2  
rank  
rank  
= NS + NU  
C1 D12  
A jωI B1  
= NS + NY  
C2 D21  
Condition 1 is a standard condition to ensure the existence of a stabilizing  
controller. Condition 2 ensures that the control signal u is contained in the  
normalized error vector e (refer to Figure 4-3). Conversely, condition 3  
ensures that some exogenous input (disturbance or noise) affects the  
measured signals (refer to Figure 4-3). Conditions 4 and 5 ensure certain  
minimal realizations of subblocks of the extended plant ([GD88]).  
[SysC,Syszw]=hinfcontr(SysAug,gamma,nw,nz,{method})  
The hinfcontr( )function designs an Hcontroller for an augmented  
plant. The augmented plant should satisfy the five restrictions in the  
Restrictions on the Extended Plant section. The hinfcontr( )function  
tests for these restrictions and returns an error if they are violated.  
Assuming the restrictions are not violated, a controller satisfying  
Hew ≤ γ will exist if certain low-level conditions also hold. These  
involve conditions for the solution of the underlying Riccati equations  
and conditions for some other constraints. The details can be found in  
[GD88,DGKF89] and are beyond the scope of this manual. If the low-level  
conditions are violated, an error statement is displayed:  
hinfcontr –>No stabilizing controller meets the spec!  
Adjust gamma and try again!  
When this occurs it means that the original gammais too small and a  
larger gamma(for example, a looser spec) is required to eliminate the  
error condition. If gammais too small, or any other necessary condition  
is not met, the hinfcontr( )function returns a zero controller and the  
closed-loop system is equal to the open-loop system:  
SysC = system( [], [], [], 0 ),  
Syszw = system(A, B1, C1, D11)  
MATRIXx Xmath Robust Control Module  
4-8  
ni.com  
   
Chapter 4  
Controller Synthesis  
If no error message occurs, then Hew ≤ γ is guaranteed. However,  
this does not preclude the possibility that either Hew « γ or that  
γ
opt « Hew .  
For the former case, there are two checks:  
Use the linfnorm( )function to compute Hew  
.
Compute the graph σmax[Hew(jω)] versus ω.  
If Hew « γ by about 6 dB or more, then you can decrease gammaand try  
again.  
When gammais very large, the specification (Equation 4-1) is easily  
met. In this case, the hinfcontr( )function returns a controller that  
approximately minimizes the H2 norm of Hew while satisfying  
Equation 4-2. Gammacan be interpreted as a “knob” that smoothly  
transforms the H2 optimal (LQG) controller, (with gammalarge), to a  
Hoptimal controller (with gamma≈ γopt ).  
Similarly, for a large gamma, the controller has good RMS performance  
with the noise spectra determined by the weights Wdist and Wnoise. For a  
small gamma, the controller has good worst-case performance for noise  
spectra that lay below the weights Wdist and Wnoise  
.
Example 4-1  
Example of hinfcontr( )  
Referring to Figure 4-2, suppose G has the state space description,  
·
x = x + d + u  
y = x + n  
where:  
x
d
n
z =  
v =  
u
1. The extended system matrix for G is:  
A = 1;  
B1 = [1,0]; B2 = 1; B = [B1,B2];  
C1 = [1;0]; C2 = 1; C = [C1;C2];  
D11 = zeros(2,2); D12 = [0;1]; D21 = [0,1]; D22 = 0;  
D = [D11,D12; D21,D22];  
G = system(A,B,C,D);  
nw = 2; nz = 2;  
© National Instruments Corporation  
4-9  
MATRIXx Xmath Robust Control Module  
 
Chapter 4  
Controller Synthesis  
Suppose the input/output weights are as follows:  
1
s + 1  
-----------  
0
1
0
0
Win  
=
Wout  
=
0.1  
0
0.1  
2. Create the four weights:  
Wdist = 1/makepoly([1,1],"s")  
Wdist (a transfer function) =  
1
-----  
s + 1  
Wnoise = 0.1;  
Wreg=1/makepoly(1,"s");  
Wact = 0.1;  
3. Combine the weights in Win and Wout (refer to Figure 4-4):  
Win = [Wdist,0,0;0,Wnoise,0;0,0,1];  
Wout=[Wreg,0,0;0,Wact,0;0,0,1];  
The resulting system, Pcan be obtained by putting in series the plant  
G and the two weights:  
P = Wout*G*Win;  
Before using the hinfcontr( )function, you must decide on an  
initial guess for gamma.  
Win  
Wout  
d
n
u
u
u
u
y
y
y
1
s + 1  
1
G
0.1  
1
0.1  
1
Figure 4-4. Plant and Weights for hinfcontr( )  
MATRIXx Xmath Robust Control Module  
4-10  
ni.com  
 
Chapter 4  
Controller Synthesis  
4. For this example, you will start with gamma=1as the initial guess and  
enter:  
[K,Hew] = hinfcontr(P,1,2,2);  
No error messages are reported. This means that a stabilizing  
controller has been found such that Equation 4-1 holds. That is,  
Hew 1 .  
The actual Hnorm is found from:  
normHew=linfnorm(Hew)  
normHew (a scalar) = 0.211984  
The result is that on this first iteration:  
gamma = 1 Æ normHew = 0.212  
Continuing to iterate for the optimal gamma:  
[K,Hew] = hinfcontr(P,.2,2,2);  
normHew=linfnorm(Hew)  
normHew (a scalar) = 0.173218  
[K,Hew] = hinfcontr(P,.15,2,2);  
normHew=linfnorm(Hew)  
normHew (a scalar) = 0.147418  
[K,Hew] = hinfcontr(P,.13,2,2);  
normHew=linfnorm(Hew)  
normHew (a scalar) = 0.13103  
[K,Hew] = hinfcontr(P,.12,2,2);  
normHew=linfnorm(Hew)  
normHew (a scalar) = 2.02252  
hinfcontr --> No stabilizing controller meets the  
spec.!!  
Adjust gamma and try again  
The iterations establish that γopt lies between 0.12 and 0.13. Figure 4-5  
shows the output of the perfplotsfunction on the closed-loop  
system Hew for γ = 0.13.  
[K,Hew] = hinfcontr(P,.13,2,2);  
normHew = linfnorm(Hew, {tol=1e-3})  
normHew (a scalar) = 0.129863  
svHew = perfplots(Hew,1,1,{Fmin=0.01,Fmax=100});  
© National Instruments Corporation  
4-11  
MATRIXx Xmath Robust Control Module  
Chapter 4  
Controller Synthesis  
Figure 4-5. Perfplots for Hew  
It also is useful to perform perfplots( )on the unweighted  
closed-loop system, Hzv, which in this case is the closed-loop transfer  
matrix from (d,n) into (x,u).  
The following function calls produce Figure 4-6:  
Hzv=clsys(G,K);  
Hzv=perfplots(Hzv,1,1,{vF=domain(svHew)});  
MATRIXx Xmath Robust Control Module  
4-12  
ni.com  
 
Chapter 4  
Controller Synthesis  
Figure 4-6. Perfplots for Hzv  
singriccati( )  
[P, solstat] = singriccati(A,Q,R {method})  
The singriccati( )function solves the Indefinite Algebraic Riccati  
Equation (ARE):  
AP + PA PRP + Q = 0  
The ARE is solved by decomposing the Hamiltonian:  
A R  
Q A′  
The required decomposition of the Hamiltonian can be achieved using  
method="eig"or method="schur". The Schur decomposition is  
slower, but might handle some ill-conditioned problems. After solving the  
ARE, singriccati( )calculates the residue (ATP + PA PRP + Q).  
A warning is displayed if the residue is large. For an example of this  
function, refer to the Xmath Help.  
© National Instruments Corporation  
4-13  
MATRIXx Xmath Robust Control Module  
       
Chapter 4  
Controller Synthesis  
Linear-Quadratic-Gaussian Control Synthesis  
The H2 Linear-Quadratic-Gaussian (LQG) control design methods are  
based on minimizing a quadratic function of state variables and control  
inputs. Conventionally, the problem is specified in the time domain.  
By converting the LQG performance index into the frequency domain,  
it becomes obvious that the conventional LQG places equal penalty on  
states and control inputs at all frequencies. It is possible to realize  
significant improvement in robustness and performance by making the  
penalty weighting matrices functions of frequency.  
LQG Frequency Shaping  
Bryson’s rule [BH69] can be extended to initially select a frequency  
shaping for a particular problem. For the control design problem, the  
frequency-shaped weighting matrices should be large at frequencies where  
control inputs are less desirable. For example, a large weighting on control  
signals at high frequency would produce less control activity at those  
frequencies, leading to a closed-loop system with lower bandwidth. Similar  
ideas apply to selection of state weighting in control design and the  
development of robust state estimators.  
Three functions are available to solve the problem of frequency-shaped cost  
functionals:  
fsregu( )—frequency-shaped regulator  
fsesti( )—frequency-shaped estimator  
fslqgcomp( )—frequency-shaped linear-quadratic-gaussian  
compensator  
fsregu( )  
[SysC, SysCC, vEV] = fsregu(SysA, ns, RXXA, RUUA, {RXUA})  
The fsregu( )function computes a frequency-shaped control law.  
It assumes you start with:  
·
x = Ax + Bu  
and  
1
2
[x*(jω)Q(jω)x(jω) + u*(jω)R(jω)u(jω)]dω  
--  
J =  
∞  
MATRIXx Xmath Robust Control Module  
4-14  
ni.com  
             
Chapter 4  
Controller Synthesis  
This expression can be converted into the following form [Gu80]:  
·
B1  
B2  
x
x
A11 A12  
A21 A22  
=
+
v
·
x′  
x′  
Ac  
Bc  
x
u =  
+ D v  
0 C12  
x′  
Cc  
Dc  
If R(jω) is not a function of frequency, then C12 = 0 and D = I.  
Note The system has a new input v and the old input u is now the output of the system.  
This structure is only used for computational convenience.  
Define:  
x
xA  
=
x'  
Thus,  
·
xA = AcxA + Bcv  
u = CcxA + Dcv  
1
2
(xA* R xA + v*Ruu v + 2xA* Rxu v)dω  
--  
J =  
xxA  
A
A
After the system is put in this form, you are ready to use the fsregu( )  
function. For more information on the fsregu( )syntax, refer to the  
Xmath Help.  
© National Instruments Corporation  
4-15  
MATRIXx Xmath Robust Control Module  
Chapter 4  
Controller Synthesis  
fsesti( )  
[SysF, vEV] = fsesti(SysA, ns, QWWA, QVVA, {QWVA})  
The fsesti( )function computes a frequency-shaped state estimator.  
The estimation problem is stated as follows:  
·
x = Ax + Bu + w  
y = Cx + Du + v  
The frequency-shaped filter design problem is to minimize,  
1
2
(w*Qww(jω)w + v*Qvv(jω)v)dω  
--  
J =  
∞  
which can be written as:  
or it can be written as:  
·
xA = AexA + Beu + wA  
·
x
x
B
A11 A12  
A21 A22  
=
+
u + wA  
·
x′  
x′  
0
xA  
Be  
Ae  
·
x
u =  
+ Deu + vA  
C1 C2  
·
x′  
Ce  
1
2
wA* Qww wA + v*AQvv vA + 2wA* Qwv v)dω  
--  
J =  
A
A
A
∞  
For more information on the fsesti( )syntax, refer to the Xmath Help.  
MATRIXx Xmath Robust Control Module  
4-16  
ni.com  
     
Chapter 4  
Controller Synthesis  
fslqgcomp( )  
[SysCC, vEV] = fslqgcomp(SysF, SysC)  
The fslqgcomp( )function combines filter and control law to compute a  
controller from a control law and an estimator. For more information on the  
fslqgcomp( )syntax, refer to the Xmath Help.  
Frequency-Shaped Control Design Commands  
This extended example uses the previously discussed functions to  
demonstrate frequency-shaped control design techniques. A four-state,  
poorly damped system is studied to demonstrate how robustness can be  
attained using frequency shaping. The control law and filter are designed  
on a reduced second order system with and without frequency shaping.  
1. Create a full-order system:  
a=[0,1,0,0;-1,-.01,0,0;0,0,0,1;0,0,-25,-.05];  
b=[0,1,0,1]';  
c=[0,1,0,-1]; d=0;  
Sys=system(a,b,c,d);  
2. Calculate open-loop eigenvalues:  
eig(a)  
ans (a column vector) =  
-0.005 + 0.999987 j  
-0.005 - 0.999987 j  
-0.025 + 4.99994 j  
-0.025 - 4.99994 j  
3. Create a reduced order system by selecting only the first mode:  
ar=a(1:2,1:2);br=b(1:2);cr=c(1:2);  
Sysr=system(ar,br,cr,d);  
4. Design an LQG compensator for the reduced-order system, without  
using frequency shaping:  
qxx=diagonal([0,1]);quu=1;  
kr=regu(Sysr,qxx,quu);  
ke=esti(Sysr,qxx,quu);  
# Linear-Quadratic-Regulator  
# Linear-Quadratic-Estimator  
Sysc=lqgcomp(Sysr,kr,ke); # LQG Compensator  
Syscl=feedback(Sysr,Sysc); # Reduced-order closed-loop  
poles(Syscl)  
ans (a column vector) =  
-0.500025 + 0.866011 j  
-0.500025 - 0.866011 j  
© National Instruments Corporation  
4-17  
MATRIXx Xmath Robust Control Module  
     
Chapter 4  
Controller Synthesis  
-0.500025 + 0.866011 j  
-0.500025 - 0.866011 j  
5. Try the LQG compensator with the full-order system:  
[Syscl_fo]=feedback(Sys,Sysc);  
poles(Syscl_fo)  
ans (a column vector) =  
-0.401519 + 0.864869 j  
-0.401519 - 0.864869 j  
-0.638796 + 0.855861 j  
-0.638796 - 0.855861 j  
0.0152647 + 4.90994 j  
0.0152647 - 4.90994 j  
6. Find a stabilizing reduced order controller using frequency shaping.  
To stabilize this system, try a frequency-shaped control-law, where the  
weighting on the control signal will be 1 + (ω)4.  
First, form an augmented system, defining two additional states,  
·
μ and u.  
The augmented system will be:  
0
0
0
0
0
1
xr  
u
xr  
u
Ar Br  
d
dt  
----  
=
+
v = AaxA + Bav  
(4-3)  
(4-4)  
0 0 0 1  
0 0 0 0  
·
·
u
u
xr  
u =  
+ (0)v = CaxA + Dav  
0 0 1 0  
u
·
u
Note The augmented system has two extra states implementing the frequency shaping on  
the control system.  
aa=[ar,br,[0;0];0,0,0,1;0,0,0,0];  
bb=[0,0,0,1]';cc=[0,0,1,0];dd=0;  
Sysa=system(aa,bb,cc,dd)  
Sysa (a state space system) =  
A
0
1
0
1
0
0
-1  
-0.01  
MATRIXx Xmath Robust Control Module  
4-18  
ni.com  
 
Chapter 4  
Controller Synthesis  
0
0
0
0
0
0
1
0
B
0
0
0
1
C
0
D
0
0
1
0
X0  
0
0
0
0
System is continuous  
7. Frequency-weight the control signal.  
Transfer the weight on Ufrom RUUto the third diagonal entry in RXXA.  
Note In Equation 4-3, u is the third state of the augmented system. RUUweighs the new  
frequency-shaped control signal v.  
Design a frequency-shaped regulator:  
rxxa=diagonal([0,1,1,0]);ruua=1;  
[Sysfs_sr,,fs_evr]=fsregu(Sysa,2,rxxa,ruua)  
Sysfs_sr (a state space system) =  
A
0
1
-1.95171  
-1.97571  
B
0
0
0.951712  
-0.228069  
C
1
0
0
D
0
X0  
0
0
© National Instruments Corporation  
4-19  
MATRIXx Xmath Robust Control Module  
Chapter 4  
Controller Synthesis  
System is continuous  
fs_evr (a column vector) =  
-0.645263 + 0.587929 j  
-0.645263 - 0.587929 j  
-0.347592 + 1.09155 j  
-0.347592 - 1.09155 j  
8. Calculate the frequency-shaped estimator:  
Sysaf=system(ar,br,cr,0);qwwa=qxx;qvva=quu;  
[Sysfs_se,fs_eve]=fsesti(Sysaf,2,qwwa,qvva)  
Sysfs_se (a state space system) =  
A
0
1
-1  
-1.00005  
B
5.52357e-17  
0.99005  
0
1
C
1
0
0
1
D
0
0
0
0
X0  
0
0
System is continuous  
fs_eve (a column vector) =  
-0.500025 + 0.866011 j  
-0.500025 - 0.866011 j  
The compensator should be structured as shown in Figure 4-7.  
x
u
A
y
Sysfs_se  
Sysfs_se  
Figure 4-7. Frequency-Shaped Compensator  
The fslqgcomp( )function can be used to develop the compensator.  
MATRIXx Xmath Robust Control Module  
4-20  
ni.com  
 
Chapter 4  
Controller Synthesis  
9. Design the LQG compensator.  
[Sysfs_sc,fs_evc]=fslqgcomp(Sysfs_se,Sysfs_sr)  
Sysfs_sc (a state space system) =  
A
0
-1  
0
1
0
0
-1.00005  
0
1
0
0
1
0.951712  
-0.228069  
-1.95171  
-1.97571  
B
5.52357e-17  
0.99005  
0
0
C
0
0
1
0
D
0
X0  
0
0
0
0
System is continuous  
fs_evc (a column vector) =  
-0.373302  
-1.17564  
-0.713411 + 1.33028 j  
-0.713411 - 1.33028 j  
Enforce negative feedback:  
Sysfs_sc = -Sysfs_sc;  
© National Instruments Corporation  
4-21  
MATRIXx Xmath Robust Control Module  
Chapter 4  
Controller Synthesis  
10. Compute the closed-loop system for the reduced order plant and the  
frequency-shaped compensator:  
[Sysfs_scl]=feedback(Sysr,Sysfs_sc);  
poles(Sysfs_scl)  
ans (a column vector) =  
-0.645263 + 0.587929 j  
-0.645263 - 0.587929 j  
-0.500025 + 0.866011 j  
-0.500025 - 0.866011 j  
-0.347592 + 1.09155 j  
-0.347592 - 1.09155 j  
11. Compute the closed-loop system for the full-order plant and the  
frequency-shaped compensator.  
Sysfs_scl_fo = feedback(Sys,Sysfs_sc);  
poles(Sysfs_scl_fo)  
ans (a column vector) =  
-0.690216 + 0.522898 j  
-0.690216 - 0.522898 j  
-0.419783 + 0.892632 j  
-0.419783 - 0.892632 j  
-0.381722 + 1.10668 j  
-0.381722 - 1.10668 j  
-0.0261589 + 5.00027 j  
-0.0261589 - 5.00027 j  
The full-order closed-loop system is stable. The open-loop eigenvalues  
of the unmodelled mode have not moved much, which is a sign of good  
robustness. The eigenvalue of the unmodelled mode changed from  
–.0250 5j to –0.0262 5j.  
Loop Transfer Recovery (lqgltr)  
Loop transfer recovery (LTR) is fully described in references [KS72,  
DoS79,DoS81,SA88]. The properties of the recovery pertain to the LQG  
feedback system as shown in Figure 4-8.  
The parameter ρ (rho) can be manipulated by the user to obtain loop  
transfer recovery through the regulator (lqrltr) or the estimator  
(lqeltr).  
MATRIXx Xmath Robust Control Module  
4-22  
ni.com  
   
Chapter 4  
Controller Synthesis  
Plant  
G(s)  
u
y
x = Ax + Bu  
x
y = Cx  
Estimator  
KR  
x = Ax + Bu + KEε  
ε = y – Cx  
x
K(s)  
Figure 4-8. LQG Feedback System for Loop Transfer Recovery  
lqgltr( )  
[SysC,EV,Kr] = lqgltr(Sys,Wx,Wy,K,rho,{keywords})  
The lqgltr( )function designs an estimator or regulator which recovers  
loop transfer robustness through the design parameter ρ (rho). For a  
discussion of the syntax and a full listing of keywords, refer to the Xmath  
Help.  
The keyword recoverspecifies whether the recovery should be achieved  
through regulator or estimator design.  
If the keyword is set to recover="regulator", the loop-transfer is  
recovered by designing a regulator with the following model:  
·
x = Ax + w  
y = Cx + v  
and the objective function:  
J =  
(x'Qx + u'Ru)dt  
1
2
--  
0
with:  
Q = Rxx + ρC'C  
R = Ruu  
© National Instruments Corporation  
4-23  
MATRIXx Xmath Robust Control Module  
     
Chapter 4  
Controller Synthesis  
Then ρ is increased so that pointwise in s:  
K(s)G(s) → KR(sI A)1B  
Regulator recovery is only guaranteed if G(s) is minimum-phase and there  
are at least as many control signals u as measurements y.  
If recover="estimator", the loop-transfer is recovered by designing an  
estimator with the following model:  
·
x = Ax + w  
y = Cx + v  
where w and v have noise intensities:  
W = Qxx + ρBB'  
V = Qyy  
Then ρ is increased so that pointwise in s:  
G(s)K(s) → C(sI A)1KE  
Recovery is only guaranteed if G(s) is minimum-phase and there are at least  
as many measurements y as there are control signals u.  
If graphis on(default) the singular value loop transfer response is plotted.  
The solid line represents the original loop transfer, and the dashed line  
represents the loop transfer after recovery.  
MATRIXx Xmath Robust Control Module  
4-24  
ni.com  
A
Bibliography  
[BBK88]  
[BeP79]  
[BoB90]  
S. Boyd, V. Balakrishnan, and P. Kabamba. “A bisection method for computing the  
L norm of a transfer matrix and related problems.” Mathematical Control Signals,  
Systems Vol. 2, No. 3, pp 207–219, 1989.  
A. Berman and R.J. Plemmons. “Nonnegative Matrices in the Mathematical  
Sciences.” Computer Science and Applied Mathematics Series, Academic  
Press, 1979.  
S. Boyd and V. Balakrishnan. “A regularity result for the singular values of a transfer  
matrix and a quadratically convergent algorithm for computing its Lnorm.”  
Systems Control Letters, Vol. 15, pp 1–7, 1990.  
[BoB91]  
[BH69]  
S. Boyd and C. Barratt. Linear Controller Design: Limits of Performance.  
Prentice-Hall, 1991.  
A.E. Bryson and Y.C. Ho. Applied Optimal Control, p 149. Blaisdell  
Publishing Co., 1969.  
[DoS79]  
[DoS81]  
J.C. Doyle and G. Stein. “Robustness with Observers.” IEEE Transactions on  
Automatic Control, August 1979.  
J.C. Doyle and G. Stein. “Multivariable Feedback Design: Concepts for a  
Classical/Modern Synthesis.” IEEE Transactions on Automatic Control,  
Vol. AC-26, pp 4–16, February 1981.  
[Doy82]  
J.C. Doyle. “Analysis of Feedback Systems with Structured Uncertainties.”  
IEEE Proceedings, November 1982.  
[DWS82]  
J.C. Doyle, J.E. Wall, and G. Stein. “Performance and Robustness Analysis for  
Structure Uncertainties.” Proceedings IEEE Conference on Decision and Control,  
pp 629–636, 1982.  
© National Instruments Corporation  
A-1  
MATRIXx Xmath Robust Control Module  
     
Appendix A  
Bibliography  
[FaT88]  
M.K. Fan and A.L. Tits. “m-form Numerical Range and the Computation of the  
Structured Singular Value.” IEEE Transactions on Automatic Control, Vol. 33,  
pp 284–289, March 1988.  
[FaT86]  
M.K. Fan and A.L. Tits. “Characterization and Efficient Computation of the  
Structured Singular Value.” IEEE Transactions on Automatic Control, Vol. AC-31,  
pp 734–743, August 1986.  
[Fr87]  
B. Francis. A Course in L Control Theory. Springer-Verlag,  
Berlin-New York, 1987.  
[FPGM87]  
D.S. Flamm, S. Boyd, G. Stein, and S.K. Mitter. “Tutorial Workshop on LControl  
Theory.” pre-conference workshop, Proceedings 26th IEEE Conference on Decision  
and Control, December 1988.  
[GD88]  
K. Glover and J.C. Doyle. “State-space formulae for all stabilizing controllers that  
satisfy an L norm bound and relations to risk sensitivity.” Systems and Control  
Letters, Vol. 11, pp 167–172, 1988.  
[DGKF89]  
J.C. Doyle, K. Glover, P.K. Khargonekar, and B. Francis. ‘‘State-space solutions to  
standard H2 and L control problems.” IEEE Transactions on Automatic Control,  
Vol. AC-34, No. 8, pp 831–847, August 1989.  
[Gu80]  
N.K. Gupta. “Frequency Shaping of Cost Functionals: An extension of LQG Design  
Methods.” AIAA Journal of Guidance and Control, Vol. 3, No. 6, December 1980.  
[ONR84]  
ONR/Honeywell Workshop on Advances in Multivariable Control. Lecture Notes,  
Minneapolis, MN, 1984.  
[Osb60]  
[Saf82]  
E.E. Osborne. “On Preconditioning of Matrices.” JACM, 7:338–345, 1960.  
M.G. Safonov. “Stability Margins of Diagonally Perturbed Multivariable Feedback  
Systems.” IEEE Proceedings, 129-D:251–256, November 1982.  
[SD83]  
[SD84]  
M.G. Safonov and J.C. Doyle. “Optimal Scaling for Multivariable Stability Margin  
Singular Value Computation.” Proceedings of MECO/EES 1983, Symposium, 1983.  
M.G. Safonov and J.C. Doyle. “Minimizing Conservativeness of Robust Singular  
Values.” Multivariable Control, pp 197–207, S.G. Tzafestas, editor. D. Reidel  
Publishing Company, 1984.  
[SLH81]  
M.G. Safonov, A.J. Laub, and G.L. Hartmann. “Feedback Properties of  
Multivariable Systems: The Role and Use of the Return Difference Matrix.”  
IEEE Transactions on Automatic Control, Vol. AC-26, February 1981.  
MATRIXx Xmath Robust Control Module  
A-2  
ni.com  
Appendix A  
Bibliography  
[SA88]  
[Za81]  
[KS72]  
G. Stein and M. Athans. “The LQG/LTR Procedure for Multivariable Control  
Design.” IEEE Transactions on Automatic Control, Vol. AC-32, No. 2, pp 105–114,  
February 1987.  
G. Zames. “Feedback and optimal sensitivity: model reference transformations,  
multiplicative semi-norms, and approximate inverses.” IEEE Transactions on  
Automatic Control, Vol. AC-26, pp 301–320, 1981.  
H. Kwakernaak and R. Sivan. Linear Optimal Control Systems. Wiley, 1972.  
© National Instruments Corporation  
A-3  
MATRIXx Xmath Robust Control Module  
B
Technical Support and  
Professional Services  
Visit the following sections of the National Instruments Web site at  
ni.comfor technical support and professional services:  
Support—Online technical support resources at ni.com/support  
include the following:  
Self-Help Resources—For answers and solutions, visit the  
award-winning National Instruments Web site for software drivers  
and updates, a searchable KnowledgeBase, product manuals,  
step-by-step troubleshooting wizards, thousands of example  
programs, tutorials, application notes, instrument drivers, and  
so on.  
Free Technical Support—All registered users receive free Basic  
Service, which includes access to hundreds of Application  
Engineers worldwide in the NI Discussion Forums at  
ni.com/forums. National Instruments Application Engineers  
make sure every question receives an answer.  
For information about other technical support options in your  
area, visit ni.com/servicesor contact your local office at  
ni.com/contact.  
Training and Certification—Visit ni.com/trainingfor  
self-paced training, eLearning virtual classrooms, interactive CDs,  
and Certification program information. You also can register for  
instructor-led, hands-on courses at locations around the world.  
System Integration—If you have time constraints, limited in-house  
technical resources, or other project challenges, National Instruments  
Alliance Partner members can help. To learn more, call your local  
NI office or visit ni.com/alliance.  
If you searched ni.comand could not find the answers you need, contact  
your local office or NI corporate headquarters. Phone numbers for our  
worldwide offices are listed at the front of this manual. You also can visit  
the Worldwide Offices section of ni.com/niglobalto access the branch  
office Web sites, which provide up-to-date contact information, support  
phone numbers, email addresses, and current events.  
© National Instruments Corporation  
B-1  
MATRIXx Xmath Robust Control Module  
                   
Index  
problem definition, 4-1  
restrictions on extended plant, 4-7  
weight selection, 4-5  
A
Algebraic Riccati Equation (ARE), 4-13  
H2 control synthesis, 4-14  
H2 norm, 3-4  
help, technical support, B-1  
hinfcontr( ), 4-2, 4-8  
C
clsys( ), 3-10  
D
instrument drivers (NI resources), B-1  
diagnostic tools (NI resources), B-1  
documentation  
drivers (NI resources), B-1  
L
E
Lnorm, 3-3  
linfnorm( ), 3-4  
F
frequency-shaped  
control law, 4-14  
filter design, 4-16  
fsesti( ), 4-16  
fsregu( ), 4-14  
M
N
National Instruments support and services, B-1  
NI support and services, B-1  
nomenclature, 1-2  
H
Hcontrol synthesis, 4-1  
building the plant model, 4-3  
extended transfer matrix, 4-2  
nominal closed-loop system, 2-2  
creating, 2-4  
© National Instruments Corporation  
I-1  
MATRIXx Xmath Robust Control Module  
 
Index  
nominal transfer function, 2-8  
norm  
software (NI resources), B-1  
H2, 3-4  
L, 3-3  
stability margin, 2-3, 2-10  
structured nonparametric uncertainty, 2-1  
structured singular value, 2-11  
support, technical, B-1  
O
osscale( ), 2-16  
feedback connection, 2-2  
nominal closed-loop, 2-1  
stability margin, 2-3  
P
pfscale( ), 2-16  
programming examples (NI resources), B-1  
technical support, B-1  
training and certification (NI resources), B-1  
R
reducibility, 2-17  
nominal, 2-8  
robust stability, 2-3  
uncertain, 2-2  
S
scaled singular values, 2-11  
scaling  
Optimal (Boyd), 2-15  
Osborne, 2-15  
Perron-Frobenius, 2-15  
ssv( ), 2-15  
wcgain( ), 2-19  
Web resources, B-1  
well-posed closed-loop system, 3-11  
worst-case gain, 2-8  
worst-case performance degradation, 2-8,  
2-18  
singriccati( ), 4-13  
singular value Bode plots, 3-1  
singular values, 2-11  
smargin( ), 2-4  
MATRIXx Xmath Robust Control Module  
I-2  
ni.com  

Motorola 2001 Portable Cell Phone User Manual
Kyocera K9 User Manual
KitchenAid Coffeemaker KCM111OB User Manual
HTC Desire Android 21 User Manual
GE PDW9700 User Manual
Evolution Technologies Coffeemaker PR10908000 User Manual
Epson CMD 2260 User Manual
Bunn Coffeemaker CDBCF DV User Manual
Bosch Appliances GSM Com 608 User Manual
Black & Decker SPACEMAKER ODC425 User Manual