Robust Control of Linear Quantum Systems

Similar documents
Robust Linear Quantum Systems Theory

Quantum Linear Systems Theory

Control and Robustness for Quantum Linear Systems

Feedback Control of Quantum Systems. Matt James ANU

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

Quantum Linear Systems Theory

arxiv: v1 [quant-ph] 18 Sep 2015

Semidefinite Programming Duality and Linear Time-invariant Systems

Introduction to numerical simulations for Stochastic ODEs

arxiv: v1 [quant-ph] 16 Mar 2016

COHERENT CONTROL VIA QUANTUM FEEDBACK NETWORKS

The norms can also be characterized in terms of Riccati inequalities.

arxiv: v1 [quant-ph] 13 Nov 2016

6.241 Dynamic Systems and Control

Linear Filtering of general Gaussian processes

Schrodinger Bridges classical and quantum evolution

arxiv: v1 [quant-ph] 20 Jun 2017

EL2520 Control Theory and Practice

Robust Control 9 Design Summary & Examples

Schrodinger Bridges classical and quantum evolution

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

Modern Optimal Control

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

6.241 Dynamic Systems and Control

Introduction to Quantum Linear Systems

Robust Control 5 Nominal Controller Design Continued

Quantum Dissipative Systems and Feedback Control Design by Interconnection

Quantum Dissipative Systems and Feedback Control Design by Interconnection

Unconditionally stable scheme for Riccati equation

Coherent controllers for optical-feedback cooling of quantum oscillators

Control, Stabilization and Numerics for Partial Differential Equations

Linear Quadratic Gausssian Control Design with Loop Transfer Recovery

ME 234, Lyapunov and Riccati Problems. 1. This problem is to recall some facts and formulae you already know. e Aτ BB e A τ dτ

arxiv: v3 [quant-ph] 30 Oct 2014

Week 9 Generators, duality, change of measure

Ian R. Petersen. University of New South Wales, Canberra

Eigenvalues and Eigenvectors

Stationary trajectories, singular Hamiltonian systems and ill-posed Interconnection

Lecture 5: Linear Systems. Transfer functions. Frequency Domain Analysis. Basic Control Design.

10 Transfer Matrix Models

BV functions in a Gelfand triple and the stochastic reflection problem on a convex set

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

Definite versus Indefinite Linear Algebra. Christian Mehl Institut für Mathematik TU Berlin Germany. 10th SIAM Conference on Applied Linear Algebra

Denis ARZELIER arzelier

Controls Problems for Qualifying Exam - Spring 2014

A Quantum Particle Undergoing Continuous Observation

The semantics of algebraic quantum mechanics and the role of model theory.

NLControl - package for modelling, analysis and synthesis of nonlinear control systems

Stability of Stochastic Differential Equations

Control of elementary quantum flows Luigi Accardi Centro Vito Volterra, Università di Roma Tor Vergata Via di Tor Vergata, snc Roma, Italy

arxiv: v1 [quant-ph] 3 Sep 2008

where r n = dn+1 x(t)

Stochastic Calculus. Kevin Sinclair. August 2, 2016

CONTROLLABILITY OF QUANTUM SYSTEMS. Sonia G. Schirmer

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

Linear Quadratic Zero-Sum Two-Person Differential Games

On Stochastic Adaptive Control & its Applications. Bozenna Pasik-Duncan University of Kansas, USA

Solution of Stochastic Optimal Control Problems and Financial Applications

Numerical treatment of stochastic delay differential equations

PROBABILITY: LIMIT THEOREMS II, SPRING HOMEWORK PROBLEMS

Short Course in Quantum Information Lecture 2

Decentralized LQG Control of Systems with a Broadcast Architecture

7.3 State Space Averaging!

Stochastic Differential Equations.

Here represents the impulse (or delta) function. is an diagonal matrix of intensities, and is an diagonal matrix of intensities.

Identification and Estimation of Quantum Linear Input-Output Systems

Lecture 3. Econ August 12

1 Linear Algebra Problems

ẋ n = f n (x 1,...,x n,u 1,...,u m ) (5) y 1 = g 1 (x 1,...,x n,u 1,...,u m ) (6) y p = g p (x 1,...,x n,u 1,...,u m ) (7)

SEA s workshop- MIT - July 10-14

Advanced Quantum Mechanics

Decay rates for partially dissipative hyperbolic systems

AN INTRODUCTION CONTROL +

Zeros and zero dynamics

Lecture 15: H Control Synthesis

LECTURE 7. k=1 (, v k)u k. Moreover r

Return Difference Function and Closed-Loop Roots Single-Input/Single-Output Control Systems

Regularization by noise in infinite dimensions

A Comparative Study on Automatic Flight Control for small UAV

Stochastic Processes. A stochastic process is a function of two variables:

Convergence of Particle Filtering Method for Nonlinear Estimation of Vortex Dynamics

Lyapunov-based control of quantum systems

Linear Algebra: Matrix Eigenvalue Problems

Optimal Polynomial Control for Discrete-Time Systems

Linear-Quadratic-Gaussian (LQG) Controllers and Kalman Filters

Analog Signals and Systems and their properties

Numerical methods for solving stochastic differential equations

arzelier

OPTIMAL CONTROL AND ESTIMATION

Perturbation theory for eigenvalues of Hermitian pencils. Christian Mehl Institut für Mathematik TU Berlin, Germany. 9th Elgersburg Workshop

Multivariate Generalized Ornstein-Uhlenbeck Processes

Modern Optimal Control

Suppose that we have a specific single stage dynamic system governed by the following equation:

Definition of a Stochastic Process

A NOTE ON STOCHASTIC INTEGRALS AS L 2 -CURVES

On the linear quadratic optimization problems and associated Riccati equations for systems modeled by Ito linear differential equations

Risk-Sensitive and Robust Mean Field Games

Control Design of a Distributed Parameter. Fixed-Bed Reactor

Lifted approach to ILC/Repetitive Control

Networked Sensing, Estimation and Control Systems

Transcription:

B. Ross Barmish Workshop 2009 1 Robust Control of Linear Quantum Systems Ian R. Petersen School of ITEE, University of New South Wales @ the Australian Defence Force Academy, Based on joint work with Matthew James, A.J. Shaiju, Hendra Nurdin, and Aline Maalouf.

B. Ross Barmish Workshop 2009 2 Introduction Recent developments in quantum and nano technology have provided a great impetus for research in quantum feedback control systems. In particular, it is being realized that robustness is critical in quantum feedback control systems, as it is in non-quantum feedback control systems. We present an overview of some recent results in the area of robust feedback control of linear quantum systems. We consider a class of quantum systems described by linear Heisenberg dynamics driven by quantum Gaussian noise processes, and controlled by a linear feedback controller which is also a quantum system.

B. Ross Barmish Workshop 2009 3 The most common area in which such systems arise is in the area of quantum optics; e.g., Photo courtesy of Elanor Huntington (UNSW@ADFA)

B. Ross Barmish Workshop 2009 4 Analysis of Linear Quantum Systems In the analysis of linear quantum systems, we consider noncommutative stochastic models of the form dx(t) = Ax(t)dt + [ B G ][ dw(t) T dv(t) T ] T ; dz(t) = Cx(t)dt + [ D H ][ dw(t) T dv(t) T ] T (1) where x(t) = [ x 1 (t)... x n (t) ] T is a vector of self-adjoint possibly non-commutative system variables. Also, A, B, G, C D are real matrices. The vector quantity w describes the input signals and is assumed to admit the decomposition dw(t) = β w (t)dt + d w(t) (2) where w(t) is the noise part of w(t) and β w (t) is a self adjoint process.

B. Ross Barmish Workshop 2009 5 In this quantum system, the input channel has two components, dw = β w dt + d w which represents disturbance signals, and dv, which represents additional noise sources. The process β w (t) serves to represent variables of other systems which may be passed to the system. This system may represent the closed loop quantum classical system in a quantum H control system.

B. Ross Barmish Workshop 2009 6 Definition. The above quantum stochastic system is said to be Strictly Bounded Real with disturbance attenuation g if there exists a positive operator valued quadratic form V (x) = x T Xx (where X is a real positive definite symmetric matrix) and constants λ > 0, ǫ > 0 such that V (x(t)) + t V (x(0)) + λt t > 0, 0 β T z β z (g 2 ǫ)β T wβ w + ǫx T x ds for all Gaussian states ρ for the initial variables x(0). Here we use the shorthand notation for quantum expectation over all initial variables and noises.

B. Ross Barmish Workshop 2009 7 Theorem. [Quantum Strict Bounded Real Lemma] The following statements are equivalent (i) The above quantum stochastic system is strictly bounded real with disturbance attenuation g. (ii) A is a stable matrix and C(sI A) 1 B + D < g. (iii) g 2 I D T D > 0 and there exists a positive definite matrix X > 0 : A T X + XA + C T C + ( XB + C T D) (g 2 I D T D) 1 (B T X + D T C) < 0. (iv) g 2 I D T D > 0 and the algebraic Riccati equation A T X + XA + C T C + (XB + C T D) (g 2 I D T D) 1 (B T X + D T C) = 0 has a stabilizing solution X 0. Furthermore, if these statements hold then X < X.

B. Ross Barmish Workshop 2009 8 H of linear quantum systems In the quantum H control problem we consider a plant which is described by noncommutative stochastic models of the form dx(t) = Ax(t)dt + B 0 dv(t) + B 1 dw(t) + B 2 du(t); x(0) = x; dz(t) = C 1 x(t)dt + D 12 du(t); dy(t) = C 2 x(t)dt + D 20 dv(t) + D 21 dw(t) We assume that the control input is of the form du(t) = β u (t)dt + dũ(t) where ũ(t) is the noise part of u(t).

B. Ross Barmish Workshop 2009 9 Assumptions 1. D T 12D 12 = E 1 > 0. 2. D 21 D21 T = E 2 > 0. [ ] A jωi B2 3. The matrix is full rank for all ω 0. C 1 D 12 [ ] A jωi B1 4. The matrix is full rank for all ω 0. C 2 D 21

B. Ross Barmish Workshop 2009 10 Riccati Equations Our quantum H result is stated in terms of the following pair of algebraic Riccati equations: (A B 2 E1 1 DT 12C 1 ) T X + X(A B 2 E1 1 DT 12C 1 ) +X(B 1 B1 T g 2 B 2 E1 1 B 2)X +g 2 C1 T (I D 12E1 1 DT 12 )C 1 = 0; (A B 1 D T 21E 1 2 C 2)Y + Y (A B 1 D T 21E 1 2 C 2) +Y (g 2 C T 1 C 1 C T 2 E 1 2 C 2)Y +B 1 (I D T 21E 1 2 D 21)B T 1 = 0.

B. Ross Barmish Workshop 2009 11 Controller Our quantum H control is defined by state space matrices constructed from the Riccati equations as follows: A K = A + B 2 C K B K C 2 + (B 1 B K D 21 )B T 1 X; B K = (I Y X) 1 (Y C T 2 + B 1 D T 21)E 1 2 ; C K = E 1 1 (g2 B T 2 X + D T 12C 1 ).

B. Ross Barmish Workshop 2009 12 Theorem. Necessity. If there exists a controller such that the resulting closed loop system is strictly bounded real with disturbance attenuation g, then the Riccati equations will have stabilizing solutions X 0 and Y 0 such that the matrix XY has a spectral radius strictly less than one. Sufficiency. Suppose the Riccati equations have stabilizing solutions X 0 and Y 0 such that the matrix XY has a spectral radius strictly less than one. If the controller is such that the matrices A K, B K, C K are as defined in terms of the Riccati solutions as above, then the resulting closed loop system will be strictly bounded real with disturbance attenuation g.

B. Ross Barmish Workshop 2009 13 Physical Realizability of Linear Quantum Systems In the quantum H control problem, the controller can be a classical system constructed from digital or analog electronics, a quantum system constructed from cavities, beamsplitters and phase shifters, or a combination of both. If the controller is to be a purely quantum system (coherent quantum feedback control), the question arises as to whether a given controller transfer function can be physically realized as a quantum system.

B. Ross Barmish Workshop 2009 14 Example Quantum H control system consisting of an optical cavity plant and an optical cavity controller: Photo courtesy of Hideo Mabuchi, Stanford.

B. Ross Barmish Workshop 2009 15 Physical Realizability of Quantum System Models To consider the issue of whether a state space quantum system model (controller) is physically realizable, it is most convenient to consider complex quantum systems of the form da = F 0 [ a a ] db = H 0 [ a a ] dt + G 0 [ dv dv ], dt + K 0 [ dv dv ]. We assume without loss of generality that the number of inputs equals the number of outputs.

B. Ross Barmish Workshop 2009 16 Also, we write F 0 = [ ] F 1 F 2 C n a 2n a, G 0 = [ ] G 1 G 2 C n a 2n v, H 0 = [ ] H 1 H 2 C n v 2n a, K 0 = [ ] K 1 K 2 C n v 2n v, [ ] [ ] F1 F F = 2 G1 G F2 F1, G = 2 G 2 G, 1 [ ] [ ] H1 H H = 2 K1 K, K = 2. H 2 H 1 K 2 K 1

B. Ross Barmish Workshop 2009 17 This complex linear quantum system is equivalent to a real linear quantum system of the form considered previously via the substitutions: x k = a k + a k, x n a +k = i(a k a k ). y l = b l + b l, y nv +l = i(b l b l), 1 l n v, w j = v j + v j, w nv +j = i(v j v j ), 1 j n v.

B. Ross Barmish Workshop 2009 18 This leads to the following system with self-adjoint variables and real coefficient matrices: dx = Ax dt + B dw, dy = Cx dt + D dw, where x = [x 1 x 2 x 2na ], y = [y 1 y 2 y 2nv ], A = 1 2 ΦFΦ, B = 1 2 ΦGΦ, C = 1 2 ΦHΦ, D = 1 2 ΦKΦ, [ ] [ ] I I I 0 Φ =, J =. ii ii 0 I

B. Ross Barmish Workshop 2009 19 Our notion of physical realizability is defined in terms of the preservation of certain quantum commutation relations and the equivalence of a quantum linear system to a quantum harmonic oscillator. An alternative algebraic condition is given in the following theorem. Theorem. The above complex linear quantum system is physically realizable if and only if there exists a hermitian matrix Ω = Ω such that FΩ + ΩF GJG = 0, G = ΩH J, K = I.

B. Ross Barmish Workshop 2009 20 Our main result on physical realizability relates this property to the system theory notion of a (J, J)-unitary transfer function. Consider the complex transfer function matrix corresponding to the above complex linear quantum system Ψ(s) = H(sI F) 1 G + K. Definition. A system with transfer function matrix Ψ(s) is said to be is said to be dual (J,J)-unitary if for s C : Re(s) 0. Ψ(s)JΨ ( s ) = J,

B. Ross Barmish Workshop 2009 21 Theorem. The above complex linear quantum system is physically realizable if and only if its transfer function matrix is dual (J,J)-unitary and K = I. This theorem shows the equivalence between the quantum mechanical notion of physical realizability and the frequency domain systems theory notion of dual (J, J)-unitary.