Relation (12.1) states that if two points belong to the convex subset Ω then all the points on the connecting line also belong to Ω.

Similar documents
14. MRAC for MIMO Systems with Unstructured Uncertainties We consider affine-in-control MIMO systems in the form, x Ax B u f x t

( ) 1 Comparison Functions. α is strictly increasing since ( r) ( r ) α = for any positive real number c. = 0. It is said to belong to

Ω ). Then the following inequality takes place:

Lecture 10: Bounded Linear Operators and Orthogonality in Hilbert Spaces

Technical Report: Bessel Filter Analysis

are positive, and the pair A, B is controllable. The uncertainty in is introduced to model control failures.

Math 7409 Homework 2 Fall from which we can calculate the cycle index of the action of S 5 on pairs of vertices as

Fuzzy n-normed Space and Fuzzy n-inner Product Space

f(1), and so, if f is continuous, f(x) = f(1)x.

REDUCING THE EFFECT OF UNMODELED DYNAMICS BY MRAC CONTROL LAW MODIFICATION. Eva Miklovičová, Ján Murgaš and Michal Gonos

2.2 Limits Involving Infinity AP Calculus

Lacunary Almost Summability in Certain Linear Topological Spaces

Modular Spaces Topology

Chapter 2. Asymptotic Notation

1-D Sampling Using Nonuniform Samples and Bessel Functions

Counting Functions and Subsets

MATH Midterm Solutions

Non-asymptotic sequential confidence regions with fixed sizes for the multivariate nonlinear parameters of regression. Andrey V.

CS537. Numerical Analysis and Computing

t is bounded. Thus, the state derivative x t is bounded. Let y Cx represent the system output. Then y

Strong Result for Level Crossings of Random Polynomials

Lecture 11. Solution of Nonlinear Equations - III

Strong Result for Level Crossings of Random Polynomials. Dipty Rani Dhal, Dr. P. K. Mishra. Department of Mathematics, CET, BPUT, BBSR, ODISHA, INDIA

A New Result On A,p n,δ k -Summabilty

1 (12 points) Red-Black trees and Red-Purple trees

CS321. Numerical Analysis and Computing

rad / sec min rev 60sec. 2* rad / sec s

Taylor Polynomials and Approximations - Classwork

Topic 9 - Taylor and MacLaurin Series

Lecture Outline. 2 Separating Hyperplanes. 3 Banach Mazur Distance An Algorithmist s Toolkit October 22, 2009

Al Lehnen Madison Area Technical College 10/5/2014

42 Dependence and Bases

) is a square matrix with the property that for any m n matrix A, the product AI equals A. The identity matrix has a ii

Lecture 24: Observability and Constructibility

KEY. Math 334 Midterm II Fall 2007 section 004 Instructor: Scott Glasgow

Multivector Functions

The Binomial Multi-Section Transformer

ECE Spring Prof. David R. Jackson ECE Dept. Notes 20

CHAPTER 6d. NUMERICAL INTERPOLATION

Math 4707 Spring 2018 (Darij Grinberg): midterm 2 page 1. Math 4707 Spring 2018 (Darij Grinberg): midterm 2 with solutions [preliminary version]

M5. LTI Systems Described by Linear Constant Coefficient Difference Equations

Math Solutions to homework 6

Lecture 19. Curve fitting I. 1 Introduction. 2 Fitting a constant to measured data

Integrals of Functions of Several Variables

2012 GCE A Level H2 Maths Solution Paper Let x,

MATH /19: problems for supervision in week 08 SOLUTIONS

Integrable Functions. { f n } is called a determining sequence for f. If f is integrable with respect to, then f d does exist as a finite real number

Minimization of the quadratic test function

If a subset E of R contains no open interval, is it of zero measure? For instance, is the set of irrationals in [0, 1] is of measure zero?

1. Using Einstein Summation notation, prove the identity: = A

EL2520 Control Theory and Practice

X. Perturbation Theory

The Binomial Multi- Section Transformer

Lower Bounds for Cover-Free Families

Error for power series (Day 2) YOU MAY USE YOUR CALCULATOR TO COMPUTE FRACTIONS AND OTHER SIMPLE OPERATIONS

CDS 101: Lecture 5.1 Controllability and State Space Feedback

International Journal of Mathematics Trends and Technology (IJMTT) Volume 47 Number 1 July 2017

Math 508 Exam 2 Jerry L. Kazdan December 9, :00 10:20

Born-Oppenheimer Approximation and Nonadiabatic Effects. Hans Lischka University of Vienna

ON CERTAIN CLASS OF ANALYTIC FUNCTIONS

5.6 Binomial Multi-section Matching Transformer

Mapping Radius of Regular Function and Center of Convex Region. Duan Wenxi

Lacunary Weak I-Statistical Convergence

An Algorithmist s Toolkit October 20, Lecture 11

Recursion. Algorithm : Design & Analysis [3]

CHAPTER 5 : SERIES. 5.2 The Sum of a Series Sum of Power of n Positive Integers Sum of Series of Partial Fraction Difference Method

COMP 2804 Solutions Assignment 1

Lecture 6: October 16, 2017

page Suppose that S 0, 1 1, 2.

Contents Two Sample t Tests Two Sample t Tests

7. Discrete Fourier Transform (DFT)

Asymptotic Expansions of Legendre Wavelet

It is often useful to approximate complicated functions using simpler ones. We consider the task of approximating a function by a polynomial.

Supplementary Material

Conditional Convergence of Infinite Products

Application to Random Graphs

u t u 0 ( 7) Intuitively, the maximum principles can be explained by the following observation. Recall

Different kinds of Mathematical Induction

ECE 901 Lecture 4: Estimation of Lipschitz smooth functions

Metric Space Properties

Jacobi symbols. p 1. Note: The Jacobi symbol does not necessarily distinguish between quadratic residues and nonresidues. That is, we could have ( a

distinct distinct n k n k n! n n k k n 1 if k n, identical identical p j (k) p 0 if k > n n (k)

Null Spaces, Column. Transformations. Remarks. Definition ( ) ( ) Null A is the set of all x's in R that go to 0 in R 3

Structure and Some Geometric Properties of Nakano Difference Sequence Space

Axioms of Measure Theory

Homework Set #3 - Solutions

SOME ARITHMETIC PROPERTIES OF OVERPARTITION K -TUPLES

Chapter 9 Computation of the Discrete. Fourier Transform

Chapter 8 Complex Numbers

By the end of this section you will be able to prove the Chinese Remainder Theorem apply this theorem to solve simultaneous linear congruences

A PROBABILITY PROBLEM

The state space model needs 5 parameters, so it is not as convenient to use in this control study.

CHAPTER 6c. NUMERICAL INTERPOLATION

CHAPTER 5 SOME MINIMAX AND SADDLE POINT THEOREMS

Progression. CATsyllabus.com. CATsyllabus.com. Sequence & Series. Arithmetic Progression (A.P.) n th term of an A.P.

5.6 Binomial Multi-section Matching Transformer

9. MRAC Design for Affine-in-Control MIMO Systems

CS 330 Discussion - Probability

n p (Ω). This means that the

Discrete Mathematics: Lectures 8 and 9 Principle of Inclusion and Exclusion Instructor: Arijit Bishnu Date: August 11 and 13, 2009

Transcription:

Lectue 6. Poectio Opeato Deiitio A.: Subset Ω R is cove i [ y Ω R ] λ + λ [ y = z Ω], λ,. Relatio. states that i two poits belog to the cove subset Ω the all the poits o the coectig lie also belog to Ω. Deiitio A.: Fuctio : R R is cove i λ + λ y λ + λ y, λ. Iequality. is illustated o Figue., ad it states that gaph o a cove uctio ust be located below the staight lie, which coects ay two coespodig uctio values. y z z y Stateet A.: Let R R the subset Ω = R Poo: Let Ω ay λ heeoe { } Figue.: A Cove Fuctio : be cove uctio. he o ay costat > is cove.,. he λ + ad. Sice λ λ + λ λ + λ = ad, cosequetly, Ω which copletes the poo. is cove the o

Stateet A.: Let R R Choose a costat > ad coside the subset Ω = R Ω ad assue that < : be cotiuously dieetiable cove uctio., i.e., { } R. Let is ot o the bouday o Ω. Also, let Ω. he the Ω ad assue that =, i.e., is o the bouday o ollowig iequality takes place:.3 whee = R is the gadiet vecto o evaluated at. Relatio.3 is illustated o Figue A.. It shows that the gadiet vecto evaluated at the bouday o a cove set always poits away o the set. Ω Poo: Sice is cove the Figue A.: Gadiet ad Cove Set λ + λ λ + λ o equivaletly: + λ + λ he o ay ozeo < λ : λ + λ < = akig the liit as λ yields elatio.3 ad copletes the poo. Suppose that, the tue paaete vecto, belogs to a cove set Ω

{ } Ω = R.4 Itoduce aothe cove set: { } Ω = R.5 It is obvious that Ω Ω. We ay ow deie the poectio opeato, which we shall use i the adaptive laws. o equivaletly: Po, y y, i = y, i y ad y y, i ot. y y, i > ad y > Po, y =.6 y, i ot aely,, y i.4. I the set { } bouday { = λ} Po does ot alte the vecto y i belogs to the cove set Ω deied, the poectio opeato subtacts a vecto oal to the o y so that we get a sooth tasoatio o the oigial vecto ield y o λ = to a taget to the bouday vecto ield o λ =. he poectio opeato cocept is illustated o Figue.3. { = λ} Po, y y { } Figue.3: Poectio Opeato

Usig Stateet A. ad iequality.3, we get the ollowig ipotat popety o the poectio opeato: y Po, y, i =, i ad y y = λ, i ot..7 o, equivaletly Po, y y.8 Based o.6, we ca ow deie the poectio opeato whe both atices o the sae diesios: Y ad Θ ae Y y y R Θ = R = ad.9, Y = Po, y Po, Po Θ y. Relatio. iplies that o atices the poectio opeato is deied colu-wise. Fially, we show that the tace tes i the Lyapuov uctio deivative.6 becoe sei-egative due to the usage o the Poectio Opeato i oig the adaptive laws.7. Sice all the tace tes i.6 ae siila, we oly coside the ist oe. t K K e PB Γ + Λ = K K Po K, Y Y λ = K Po K, Y K Y. Usig. yields the sae adaptatio law o K as i.7, aely: K =Γ Po K, e PB. Basically, Poectio Opeato esues that the colus ati a K do ot eceed thei pe-speciied bouds K K o the adaptive paaete, ad at the sae tie the opeato cotibutes to the egative sei-deiiteess o the Lyapuov uctio.6. 3

et we show how to deie cove uctio = { Ω },, = i ad cove sets. Both the uctio ad the set deiitios ae based o the desied uppe a bouds K K that ae iposed colu-wise. Fo a th colu K o the adaptive paaete ati K R, choose poectio toleace ε > ad deie i as i.: Usig.3, the sets a K K a ε K = K =.3 Ω ae deied as: Ω = K R.4 { } Fo.4 it ollows that o each =,, : a K R K K Ω = a Ω = K R K K + ε.5 We eed to copute the gadiet o the cove uctio.3. = K = K a ε K ε a K.6 Usig., the adaptive law o K ca be witte colu-wise as: >, i e PB + e PB e PB =Γ < e PB, i ot K.7 he adaptatio pocess i.7 esues uio boudedess o each colu o the adaptive paaete ati K owad i tie, that is: 4

a { } a { ε } K K K t K +, t,.8 he est o the adaptive paaetes i.7 ae deied i a siila ashio. 3. Poectio Opeato based MRAC Desig o MIMO Systes with Ustuctued Ucetaities Usig the adaptive laws.7, it is easy to see that the deivative o the Lyapuov uctio. satisies the ollowig iequality: i λi εa V = e Qe+ e PBΛε λ Q e + PBΛ ε e = e Q e PBΛ a 3. At the sae tie, due to Poectio Opeato popeties, all the adaptive paaetes ae UUB. Cosequetly, usig heoe 5., we coclude that the taectoies o the closedloop syste.-.5-.7 ae UUB. Moeove, the tackig eo e= etes a eighbohood o the oigi, withi a iite tie. he adius o the eighbohood i.e., the ultiate boud is deteied by the iiu level set o the Lyapuov uctio V which cotais the set E whee V : e PBΛ ε E = e R : e λi Q a { a K R : K K ε, } a : K R K K ε, + { } a { : Θ R ε, } + Θ + 3. 4. Adaptive Augetatio o a Baselie Cotolle Coside MIMO plat dyaics: = A + B Λ u+ 4. p p p p p 5

whee p R is the -diesioal syste state vecto, u R is the -diesioal vecto o idepedet vitual cotol iputs, A p is the -by- kow costat ati, B p is the -by- kow costat ati, Λ is the -by- ukow costat diagoal ati with positive diagoal eleets, ad p is the -by- ukow possibly oliea state-depedet vecto. Assuig o ucetaities i the odel, that is I, p Λ= =, a baselie oial liea dyaic cotolle ca be desiged to povide coad tackig. he cotolle dyaics ca be witte i the o: = A + B B 4. c c c c p + c c whee c R is the c diesioal cotolle state vecto, R is the diesioal eeece desied coad sigal. he coespodig augeted platcotolle syste becoes: p Ap p Bp = u p B c c A + c Λ + + c Bc A B B 4.3 he baselie cotolle is deied as: u = K + K 4.4 bl whee K R ad K R atices, coespodigly, with = p + c. he eeece odel dyaics is chose as: ae the eedback ad eedowad oial gai = A + B + B u 4.5 e e bl o equivaletly e = A+ B K e + B + B K Ae Be 4.6 whee A e Ap = B c Ac 4.7 6

It is assued that the baselie cotolle is desiged such that Assuptio 4.: I Λ= I ad p i 4.4 yields asyptotic tackig, i.e., li e t = t A e is Huwitz. = the the baselie cotol eedback u bl. he cotol obective is to id u i 4. such that the state o the augeted syste 4.3 tacks the state e o the eeece odel 4.6 i the pesece o the syste ucetaities Λ ad p, while all sigals i the closed-loop syste eai bouded. o this ed, we itoduce the tackig eo vecto: e= e 4.8 otal cotol iput is deied as adaptive augetatio o the baselie cotolle: u = K + K + K + K = u + u p bl ad ubl uad 4.9 whee K R ad K R R is ae the iceetal adaptive gais ad the estiated ucetaity. Substitutig 4.9 ito 4.3, yields: = A+ B Λ K + K + B + B Λ K + K BΛ p p 4. A B p Assuptio 4.: Give a costat diagoal ati Λ, thee eist "ideal" ati gais K R ad K R such that the ucetaity atchig coditios take place: p Ae = A+ B Λ K Be = B + BΛ K 4. ote that the kowledge o the tue gais will ot be equied. Fo 4. - 4., it iediately ollows that: 7