APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS

Similar documents
Perturbation theory of boundary value problems and approximate controllability of perturbed boundary control problems

On feedback stabilizability of time-delay systems in Banach spaces

DELAY-DIFFERENTIAL systems arise in the study

Stabilization of Distributed Parameter Systems by State Feedback with Positivity Constraints

On Controllability of Linear Systems 1

Conservative Control Systems Described by the Schrödinger Equation

Exponential stabilization of a Rayleigh beam - actuator and feedback design

Chapter 3. LQ, LQG and Control System Design. Dutch Institute of Systems and Control

Positive Stabilization of Infinite-Dimensional Linear Systems

Stability of an abstract wave equation with delay and a Kelvin Voigt damping

Average-Consensus of Multi-Agent Systems with Direct Topology Based on Event-Triggered Control

Control, Stabilization and Numerics for Partial Differential Equations

Fact Sheet Functional Analysis

Switching, sparse and averaged control

Applied Analysis (APPM 5440): Final exam 1:30pm 4:00pm, Dec. 14, Closed books.

A Smooth Operator, Operated Correctly

The Role of Exosystems in Output Regulation

Control Design of a Distributed Parameter. Fixed-Bed Reactor

Semigroup Generation

LQR, Kalman Filter, and LQG. Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin

SPECTRAL THEOREM FOR SYMMETRIC OPERATORS WITH COMPACT RESOLVENT

Linear Quadratic Zero-Sum Two-Person Differential Games Pierre Bernhard June 15, 2013

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION - Vol. VII - System Characteristics: Stability, Controllability, Observability - Jerzy Klamka

Module 07 Controllability and Controller Design of Dynamical LTI Systems

On the bang-bang property of time optimal controls for infinite dimensional linear systems

APPROXIMATE CONTROLLABILITY OF DISTRIBUTED SYSTEMS BY DISTRIBUTED CONTROLLERS

Topic # /31 Feedback Control Systems. Analysis of Nonlinear Systems Lyapunov Stability Analysis

Left invertible semigroups on Hilbert spaces.

Criterions on periodic feedback stabilization for some evolution equations

Strong stability of neutral equations with dependent delays

ECE7850 Lecture 7. Discrete Time Optimal Control and Dynamic Programming

Domain Perturbation for Linear and Semi-Linear Boundary Value Problems

Floor Control (kn) Time (sec) Floor 5. Displacement (mm) Time (sec) Floor 5.

CONTROLLABILITY AND OBSERVABILITY OF 2-D SYSTEMS. Klamka J. Institute of Automatic Control, Technical University, Gliwice, Poland

The Kalman-Yakubovich-Popov Lemma for Differential-Algebraic Equations with Applications

1. Find the solution of the following uncontrolled linear system. 2 α 1 1

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

Control Systems Design, SC4026. SC4026 Fall 2009, dr. A. Abate, DCSC, TU Delft

Analysis of Discrete-Time Systems

EQUIVALENT CONDITIONS FOR EXPONENTIAL STABILITY FOR A SPECIAL CLASS OF CONSERVATIVE LINEAR SYSTEMS

The i/s/o resolvent set and the i/s/o resolvent matrix of an i/s/o system in continuous time

An Operator Theoretical Approach to Nonlocal Differential Equations

Linear Quadratic Zero-Sum Two-Person Differential Games

Control Systems Design, SC4026. SC4026 Fall 2010, dr. A. Abate, DCSC, TU Delft

Converse Lyapunov-Krasovskii Theorems for Systems Described by Neutral Functional Differential Equation in Hale s Form

An introduction to Mathematical Theory of Control

A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems

6.241 Dynamic Systems and Control

ACM/CMS 107 Linear Analysis & Applications Fall 2016 Assignment 4: Linear ODEs and Control Theory Due: 5th December 2016

Linear System Theory

1 Math 241A-B Homework Problem List for F2015 and W2016

Chapter 4 Optimal Control Problems in Infinite Dimensional Function Space

Analysis of Discrete-Time Systems

Extensions and applications of LQ

Output Regulation of the Tigan System

Pseudospectra and Nonnormal Dynamical Systems

Semidefinite Programming Duality and Linear Time-invariant Systems

Myopic Models of Population Dynamics on Infinite Networks

UNCERTAINTY FUNCTIONAL DIFFERENTIAL EQUATIONS FOR FINANCE

P(E t, Ω)dt, (2) 4t has an advantage with respect. to the compactly supported mollifiers, i.e., the function W (t)f satisfies a semigroup law:

{σ x >t}p x. (σ x >t)=e at.

A Unified Formulation of Gaussian Versus Sparse Stochastic Processes

The ϵ-capacity of a gain matrix and tolerable disturbances: Discrete-time perturbed linear systems

ANALYTIC SEMIGROUPS AND APPLICATIONS. 1. Introduction

Bayesian inverse problems with Laplacian noise

A fast and well-conditioned spectral method: The US method

CONTROL OF DIGITAL SYSTEMS

Exponential stability of families of linear delay systems

BOUNDARY VALUE PROBLEMS IN KREĬN SPACES. Branko Ćurgus Western Washington University, USA

Decay rates for partially dissipative hyperbolic systems

Elliptic Operators with Unbounded Coefficients

Realization theory for systems biology

The Dirichlet-to-Neumann operator

Partial Differential Equations

Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science : MULTIVARIABLE CONTROL SYSTEMS by A.

Controller Design for Robust Output Regulation of Regular Linear Systems

Technische Universität Dresden

Review of Controllability Results of Dynamical System

Gramians based model reduction for hybrid switched systems

Information Structures Preserved Under Nonlinear Time-Varying Feedback

A Concise Course on Stochastic Partial Differential Equations

Delay-independent stability via a reset loop

Introduction to Instantons. T. Daniel Brennan. Quantum Mechanics. Quantum Field Theory. Effects of Instanton- Matter Interactions.

Nonlinear Dynamical Systems Lecture - 01

Strong uniqueness for stochastic evolution equations with possibly unbounded measurable drift term

CIMPA Summer School on Current Research on Finite Element Method Lecture 1. IIT Bombay. Introduction to feedback stabilization

MATH4406 (Control Theory) Unit 6: The Linear Quadratic Regulator (LQR) and Model Predictive Control (MPC) Prepared by Yoni Nazarathy, Artem

On the Stabilization of Neutrally Stable Linear Discrete Time Systems

A Globally Stabilizing Receding Horizon Controller for Neutrally Stable Linear Systems with Input Constraints 1

EL2520 Control Theory and Practice

Mathematical foundations - linear algebra

Nonlinear error dynamics for cycled data assimilation methods

Stabilization of second order evolution equations with unbounded feedback with delay

Math 4263 Homework Set 1

Discrete and continuous dynamic systems

Energy-based Swing-up of the Acrobot and Time-optimal Motion

16 1 Basic Facts from Functional Analysis and Banach Lattices

Invertibility and stability. Irreducibly diagonally dominant. Invertibility and stability, stronger result. Reducible matrices

Chap. 3. Controlled Systems, Controllability

Delay-dependent Stability Analysis for Markovian Jump Systems with Interval Time-varying-delays

Transcription:

APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES Said Hadd and Qing-Chang Zhong Dept. of Electrical Eng. & Electronics The University of Liverpool Liverpool, L69 3GJ United Kingdom Q.Zhong@liv.ac.uk

Outline Quick overview of recent activities Motivation of the research Problem statement Methodology Main results Special cases S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 2/50

Recent research activities Control engineering Power electronics: grid connection etc Renewable energy: wind power Automotive electronics: hybrid electric vehicles Control theory Robust control: J-spectral factorisation, algebraic Riccati equations etc Time-delay systems: a series of problems Infinite-dimensional systems: feedback stabilizability, controllability S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 3/50

Key publications One research monograph IEEE Trans. Automatic Control: 7 papers Automatica: 4 papers other IEEE Transactions: 3 papers S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 4/50

Current funding EPSRC: EP/C005953/1, 126k, to expire in 09/2008 EPSRC: EP/E055877/1, 88k EPSRC: one DTA studentship, 50k EPSRC and Add2 Ltd: Dorothy Hodgkin Postgraduate Award, 90k ESPRC and Nheolis, France: Dorothy Hodgkin Postgraduate Award, 90k S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 5/50

Motivation Controllability is one of the most important properties of a control system. Basically, it tells us if a system is controllable, in other words, whether a point in the state space is reachable. The most widely known formula is the Kalman rank condition Rank( B AB A n 1 B ) = n. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 6/50

Controllability: Analysis and design Normally, controllability is used to analyse whether a system is controllable. It can also be used for design. Many flexible modes to be suppressed The number of actuators to be minimised Many locations for placing actuators The question is where to put them and how many? S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 7/50

Infinite-dimensional case For finite-dimensional systems, there are different ways to define controllability. However, they all end up with the same concept. However, this is no longer true for infinite-dimensional case. Different concepts need to be used. Approximate or exact controllability in finite or infinite time etc. (Nine different concepts are defined in Stafani s book) S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 8/50

Approximate controllability (AC) For a given arbitrary ε > 0, it is possible to steer the state from the origin to the neighbourhood, with a radius ε, of all points in the state space. This is weaker than the exact controllability, where ε = 0. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 9/50

Some notation X, U, Y are Banach spaces, p > 1 is a real number and (A, D(A)) is the generator of a C 0 -semigroup (T(t)) t 0 on X. The type of T(t) is defined as ω 0 (A) := inf{t 1 log T(t) : t > 0}. The domain D(A), which is a Banach space, is endowed with the graph norm x A := x + Ax, x D(A). Denote the resolvent set of A by ρ(a) and the resolvent operator of A by R(µ,A) := (µ A) 1 for µ ρ(a). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 10/50

The completion of X with respect to the norm x 1 = R(µ,A)x for x X and some µ ρ(a) is a Banach space denoted by X 1, which is called the extrapolation space of A. For any Banach spaces E and F, denote the Banach space of all linear bounded operators from E to F by L(E,F) with L(E) := L(E, E). The history function of z : [ r, ) Z is the function z t : [ r, 0] Z defined by z t (θ) = z(t + θ) for t 0 and θ [ r, 0] with delay r > 0. For p > 1, L p ([ r, 0],Z) is the Banach space of all p- integrable functions f : [ r, 0] Z and W 1,p ([ r, 0],Z) is its associated Sobolev space. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 11/50

Problem statement To investigate the approximate controllability of d ( ) x(t) Dxt K 0 u(t) K 1 u t = B0 u(t) + B 1 u t dt +A ( ) x(t) Dx t K 0 u(t) K 1 u t + Lxt, t 0, ( ) lim x(t) Dxt K 0 u(t) K 1 u t = z0, t 0 x 0 = ϕ, u 0 = ψ. (1) Here x : [ r, ) X, u : [ r, ) U, the operators D,L : W 1,p ([ r, 0],X) X, K 1,B 1 : W 1,p ([ r, 0],U) X and B 0,K 0 : U X are linear bounded, and A is the generator of a C 0 -semigroup T := (T(t)) t 0 on X. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 12/50

How general it is? Distributed delays in the state AND the input, could be multiple discrete delays Delays in the derivative: neutral K 0 0 and K 1 0. d ( ) x(t) Dxt K 0 u(t) K 1 u t = B0 u(t) + B 1 u t dt +A ( ) ( x(t) Dx t K 0 u(t) ) K 1 u t + Lxt, t 0, lim x(t) Dxt K 0 u(t) K 1 u t = z0, t 0 x 0 = ϕ, u 0 = ψ. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 13/50

Background This area had been very active during 70s and 80s, focusing on retarded systems with state or input delays. Recent works include: 1) Diblik et al: Using a representation of solutions with the aid of a discrete matrix delayed exponential for linear discrete-time systems with state delays; 2) Sun et al: Using the matrix Lambert W function for linear (continuous-time) systems with state delays; 3) Rabah et al: Using the moment problem approach for neutral systems with distributed state delays. The closest one is the last one, where K 0 = 0, K 1 = 0, A = 0 and B 1 = 0. Also, the operator D is more general here. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 14/50

Key steps Transform the problem into a perturbed boundary control problem Solve the approximate controllability for the boundary control problem Then transfer the results back to the original problem S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 15/50

More notation Define Q X mϕ = ϕ, Q U mψ = ψ, D(Q X m) = W 1,p ([ r, 0],X), D(Q U m) = W 1,p ([ r, 0],U), and Q X ϕ = ϕ, D(Q X ) = {ϕ W 1,p ([ r, 0],X) : ϕ(0) = 0}, Q U ψ = ψ, D(Q U ) = {ψ W 1,p ([ r, 0],U) : ψ(0) = 0}. It is well known that Q X and Q U generate the left shift semigroups on L p ([ r, 0],X) and L p ([ r, 0],U), respectively. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 16/50

Denote by Q X 1 and Q U 1 the extension of Q X and Q U, respectively, and define the operators, for µ C, where β X := (µ Q X 1)e µ, β U := (µ Q U 1)e µ (2) e µ : Z L p ([ r, 0],Z), (e µ z)(θ) = e µθ z (3) is defined for µ C, z Z, and θ [ r, 0]. Denote by J Z (Z = X or U) the following continuous injection J Z : D(Q Z ) W 1,p ([ r, 0],Z). (4) S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 17/50

Transforming into a boundary control problem Introduce the space X := X L p ([ r, 0],X) L p ([ r, 0],U), and the matrix operator A L B 1 A m := 0 Q X m 0, 0 0 Q U m D(A m ) := D(A) W 1,p ([ r, 0],X) W 1,p ([ r, 0],U), S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 18/50

Moreover, define the boundary operators N, M : D(A m ) U := X U as ( ) ( ) 0 δ0 0 I D K1 N =, M =, 0 0 δ 0 0 0 0 and, for any u U, the operators Bu = ( B0 u 0 0 ), Ku = ( K 0 u u ) S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 19/50

Then the system (1) can be reformulated as ẇ(t) = A m w(t) + Bu(t), t 0, Nw(t) = Mw(t) + Ku(t). (5) by setting w(t) = (z(t),x t,u t ), t 0 with w(0) = = (z 0,ϕ,ψ), where z(t) = x(t) Dx t K 0 u(t) K 1 u t with limz(t) = z 0. Here, A m : X m X and t 0 N, M : X m U are linear operators with Banach spaces X, U and X m being a dense domain of X endowed with a norm, which is finer than the norm of X such that (X m, ) is complete. This is a boundary control problem. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 20/50

Verification of conditions Standard conditions for boundary control problems: (H1) A = A m with domain D(A) := KerN generates a C 0 -semi group (T (t)) t 0 on X, (H2) ImN = U. These assumptions show that the boundary value problem with M = 0 and B = 0 is well-posed in the sense that it can be reformulated as a well-posed open loop system on the state space X and control space U. Clearly N satisfies the condition (H2). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 21/50

In order to verify condition (H1), assume (A1) (Q X,β X,LJ X ) generates a regular linear system on L p ([ r, 0],X), X, X, (A2) (Q U,β U,B 1 J U ) generates a regular linear system on L p ([ r, 0],U), U, X. Define the operator A := A L B 1 0 Q X 0, 0 0 Q U D(A) := KerN = D(A) D(Q X ) D(Q U ). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 22/50

It satisfies A = A 0 + L with A 0 0 A 0 := 0 Q X 0, L := 0 0 Q U 0 L B 1 0 0 0 0 0 0 D(A 0 ) := KerN., D(L) := KerN. Clearly, A 0 generates a diagonal C 0 -semigroup on X. It is not difficult to see that if L and B 1 are admissible observation operators for Q X and Q U (in particular if (A1) and (A2) holds), respectively, then L is a Miyadera Voigt perturbation for A 0. Then A generates a C 0 -semigroup T := (T (t)) t 0 on X. Hence N satisfies the condition (H1). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 23/50

Admissible observation operator Definition 1. C L(D(A),Y ) is called an admissible observation operator for A if, for constants τ 0 and γ := γ(τ) > 0, τ 0 CT(t)x p dt γ p x p, x D(A), (6) The map Ψ x := CT( )x, defined on D(A), extends to a bounded operator Ψ : X L p loc (R +,Y ). For any x X we set Ψ(t)x := Ψ x on [0,t] and call (T(t), Ψ(t)) t 0 an observation system. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 24/50

Admissible control operator Definition 2. B L(U,X 1 ) is called an admissible control operator for (A, D(A)), the generator of a C 0 - semigroup (T(t)) t 0 on X, if for all t > 0 and u L p ([ r, 0],U) the control map Φ(t)u := takes values in X. t 0 T(t s)bu(s)ds (7) The pair (T, Φ) is called a control system represented by the operator B. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 25/50

Formal definition of AC Definition 3. Let (T, Φ) be the control system represented by the operator B. Define the reachability space R := t 0 RanΦ(t). Then, (A, B) is said to be approximately controllable if R is dense in X. This means that, for a given arbitrary ε > 0, it is possible to steer the state from the origin to the neighbourhood, with a radius ε, of all points in the state space. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 26/50

Further transformation It can be proved that the boundary control system (5) is equivalent to the distributed-parameter system ẇ(t) = Aw(t) + (B + BK)u(t), w(0) =, (8) for t 0, with A and B defined as Ax = A m x, D(A) = { x D(A m ) : Nx = Mx }. B := (µ A 1 )D µ, µ ρ(a). where D µ is the Dirichlet operator D µ := ( N Ker(µ Am )) 1 : U Ker(µ Am ). (9) S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 27/50

AC for the boundary control problem When the conditions (H1) and (H2) are satisfied, D µ exists and is bounded. For µ ρ(a), there is (A m A 1 ) = (µ A 1 )D µ N on D(A m ). Definition 4. Assume that (H1) (H2) are satisfied. The boundary value problem (5) is said to be approximately controllable if the open-loop system (A, B+BK) is (in the sense of Definition 3). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 28/50

Theorem Theorem 5. Assume that (H1) (H2) are satisfied. The boundary value problem (5) is approximately controllable if and only if, for µ ρ(a) ρ(a) and ϕ X, the fact that (I D µ M) 1 (D µ K + R(µ, A)B)u,ϕ = 0, for u U, (10) implies that ϕ = 0. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 29/50

Going back to the neutral system... S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 30/50

Dirichlet operator associated with N Note that ρ(a) = ρ(a). For µ ρ(a), one can see that the Dirichlet operator associated with N, as defined in (9), is given by ( D uv ) R(µ,A)(Le µ u + B 1 e µ v) ( µ = e µ u, uv ) U. e µ v (11) S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 31/50

More assumptions Let J : D(A) D(A) W 1,p ([ r, 0],X) W 1,p ([ r, 0],U) be the continuous injection, and set C := MJ L(D(A), U). Then C is an admissible observation operator for A under the following assumptions: (A3) (Q X,β X,DJ X ) generates a regular linear system on L p ([ r, 0],X), X, X with I X as an admissible feedback, (A4) (Q U,β U,K 1 J U ) generates a regular linear system on L p ([ r, 0],U), U, X. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 32/50

(µ) Define (µ) := I De µ R(µ,A)Le µ, µ ρ(a). By using (11), for µ ρ(a), we have ( ) (µ) (K1 e I U MD µ = µ + R(µ,A)B 1 e µ ). 0 I U (12) Thus, (µ) is invertible if and only if µ ρ(a) ρ(a). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 33/50

AC: the transformed system Definition 6. Assume that the conditions (A1) to (A4) are satisfied. Define R(t)u := t 0 (T I ) 1 (t τ)(bk + B)u(τ)dτ for t 0 and u L p loc (R +,U). The system (5) is said to be Υ-approximately controllable if Cl( t 0 P Υ (RanR(t))) = Υ, where Υ can be any of X, X, X L p ([ r, 0],X), L p ([ r, 0],X) and L p ([ r, 0],U), and P Υ is the projection operator from X to Υ. In particular, P X = I. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 34/50

w(t) = z(t) x t u t X-approximately controllable: the state z is approximately controllable; X L p ([ r, 0],X)-approximately controllable: states z and x t are approximately controllable; L p ([ r, 0],U)-approximately controllable: state u t is approximately controllable. Apparently, it is always L p ([ r, 0],U)- approximately controllable. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 35/50

AC: Transformed system Theorem 7. Assume that (A1) to (A4) are satisfied. Then the following conditions are equivalent: (i) the system (5) is X -approximately controllable, (ii) it is X L p ([ r, 0],X)-approximately controllable, (iii) for µ ρ(a) ρ(a), ϕ L q ([ r, 0],X) with 1 p + 1 q = 1 and x X, the fact that, for u U, (Ω(µ) + Γ(µ)Λ(µ))u,x + (eµ K 0 + e µ (µ) 1 Λ(µ))u,ϕ = 0 implies that x = 0 and ϕ = 0. Here, Γ(µ) := R(µ,A)Le µ (µ) 1, Ω(µ) := R(µ,A)(Le µ K 0 + B 1 e µ + B 0 ), Λ(µ) : = Ω(µ) + De µ K 0 + K 1 e µ. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 36/50

Outline of proof (I D µ M) 1 = I + Γ(µ) Γ(µ)D Γ(µ)K 1 e µ (µ) 1 I + e µ (µ) 1 D e µ (µ) 1 K 1. 0 0 I R(µ, A) = R(µ, A) R(µ, A)LR(µ, Q X ) R(µ, A)B 1 R(µ, Q U ) 0 R(µ, Q X ) 0. 0 0 R(µ, Q U ) D µ Ku + R(µ, A)Bu = Ω(µ)u e µ K 0 u. e µ u Hence, Ω(µ) + Γ(µ)Λ(µ) (I D µ M) 1 (D µ Ku + R(µ, A)Bu) = e µ K 0 + e µ (µ) 1 Λ(µ) u S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 37/50 e µ

State z(t) Assume that the conditions (A1) to (A4) are satisfied. Then the system (5) is X-approximately controllable iff, for µ ρ(a) ρ(a) and x X, the fact that (Ω(µ) + Γ(µ)Λ(µ))u,x = 0, for u U, implies that x = 0. This characterises the condition when z(t), partial state of system (5), can reach all points in X. In general, this is irrelevant to the approximate controllability of system (1). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 38/50

State x t Assume that the conditions (A1) to (A4) are satisfied. Then the system (5) is L p ([ r, 0],X)-approximately controllable iff, for µ ρ(a) ρ(a) and ϕ L q ([ r, 0],X) with 1 p + 1 q = 1, the fact that eµ (K 0 + (µ) 1 Λ(µ))u,ϕ = 0, for u U, (13) implies that ϕ = 0. This actually describes the approximate controllability of system (1). S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 39/50

AC: the neutral system Theorem 8. Assume that the conditions (A1) to (A4) are satisfied. Then the general neutral system (1) is approximately controllable iff, for µ ρ(a) ρ(a) and x X, the fact that (K0 + (µ) 1 Λ(µ))u,x = 0, for u U, implies that x = 0. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 40/50

Special case I: Difference equations with state and input delays Consider the difference equation x(t) = Dx t + K 0 u(t) + K 1 u t, t 0, (14) with x 0 = ϕ, u 0 = ψ. Note that the equation (14) is a special case of (1), with A = 0, L = 0,B 0 = 0, B 1 = 0. In this case, (µ) = I De µ, Γ(µ) = 0, Ω(µ) = 0, Λ(µ) = De µ K 0 + K 1 e µ. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 41/50

Define the operator Q I ϕ = ϕ, { } D(Q I ) = ϕ W 1,p ([ r, 0],X) : ϕ(0) = Dϕ. (µ) is invertible if and only if µ ρ(q I ). Assume that the conditions (A3) to (A4) are satisfied. Then the system (14) is approximately controllable iff, for µ ρ(q I ) and x X, the fact that (I Deµ ) 1 (K 0 + K 1 e µ )u,x = 0, for u U implies that x = 0. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 42/50

Special case II: Retarded delay systems with state and input delays Consider the following retarded delay system with finite-dimensional delay-free dynamics, which widely exist in engineering: ẋ(t) = Ax(t)+L r x(t r)+b 0 u(t)+b r u(t r), t 0, where x R n, A R n n, L r R n n, B 0 R n m, B r R n m and u R m. The variants of this system have been targets of most of the papers in this area. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 43/50

This system can be obtained from the system (1) by taking D = 0, K 0 = 0, K 1 = 0, L = L r δ r and B 1 = B r δ r, where δ r is the Dirac operator. These operators satisfy the conditions (A1) (A4). Then, (µ) = I e rµ R(µ,A)L r, µ ρ(a). Define ρ = {µ : (µ) is invertible}. Then for µ ρ(a) ρ, Γ(µ) = e rµ R(µ,A)L r (µ) 1 = (µ) 1 I, Λ(µ) = Ω(µ) = R(µ,A)[e rµ B r + B 0 ]. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 44/50

Now Theorem 8 shows that the system is approximately controllable iff, for µ ρ(a) ρ and x R n, the fact that (µ) 1 Λ(µ)u,x = 0, for u U, (15) implies that x = 0. It can be found that the equality in (15) is equivalent to that (µi A e rµ L r ) 1 (e rµ B r + B 0 )u,x = 0. When B r = 0, this condition is simpler than the one for systems with state delays obtained by Manitius et al in 70s. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 45/50

A rank condition Assume L r = 0. The left-hand side of the equality in (15), assuming µ 0 for the moment, is u (e rµ B r + B 0 ) (µi A) x = u (e rµ B r + B 0 ) Σ n 1 k=0 f k( 1 µ )(Ak ) x = Σ n 1 k=0 u (f k ( 1 µ )e rµ B r + f k ( 1 µ )B 0) (A k ) x where (µi A) 1 is replaced with (µi A) 1 = 1 µ (I 1 µ A) 1 = Σ n 1 k=0 f k( 1 µ )Ak with f k ( 1 µ ), k = 0, 1,n 1, as polynomials in 1 µ. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 46/50

This can be written as uf 0 ( 1 µ )e rµ uf 1 ( 1 µ )e rµ. uf n 1 ( 1 µ )e rµ uf 0 ( 1 µ ) uf 1 ( 1 µ ). uf n 1 ( 1 µ ) B r B r A. Br (A ) n 1 B0 B0 A. B 0 (A ) n 1 2n m n x = 0. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 47/50

Hence, the system is approximately controllable if and only if Rank ( B r AB r A n 1 B r B 0 AB 0 A n 1 B 0 ) = n. The condition µ ρ(a) ρ and µ 0 are now irrelevant. This recovers the results for systems with input delays by Sebakhy (1971). If B r = 0, then Rank ( B 0 AB 0 A n 1 B 0 ) = n. We have not lost ourselves! S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 48/50

A rank condition when L r 0 can be derived similarly. Instead of using A k in the series expansion, (A + e rµ L r ) k, k = 0, 1,n 1, should be used. However, it is difficult to express the condition for the general case in a compact form and hence omitted. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 49/50

Summary The approximate controllability of neutral systems is discussed via converting the problem into that for a perturbed control problem. After solving the approximate controllability for the perturbed control problem, the conditions for the approximate controllability of neutral systems are obtained. Some special cases are discussed. S. HADD AND Q.-C. ZHONG: APPROXIMATE CONTROLLABILITY OF NEUTRAL SYSTEMS WITH STATE AND INPUT DELAYS IN BANACH SPACES p. 50/50