Optimal control of quantum systems using dynamic programming

This thesis focuses on the optimal control of a class of closed 1 quantum systems. It encompasses and addresses the following themes 1. The theoretical underpinnings of the dynamic programming method for quantum spin systems. 2. The efficient numerical implementation of algorithms to generate optimal control laws. The use of the dynamic programming principle from optimal control theory for the control of quantum systems is introduced, while taking into account the properties of the manifolds on which they evolve. This enables the unified application of dynamic programming to disparate problems which were previously approached using distinct tools. Moreover the approach herein is applicable, in principle, to an arbitrarily large quantum system - thereby overcoming the limitations present in some other commonly used methods in the literature, on the size of the quantum system being analyzed. As the optimal cost function is non-differentiable in several cases of interest, this thesis provides a rigorous interpretation of this cost function in terms of viscosity solutions of the associated Hamilton-Jacobi-Bellman (HJB) partial differential equation (PDE) over the manifolds of interest. This framework is applied to explore optimal control strategies for drift and drift-less quantum systems that arise in various applications - thus demonstrating the efficacy of a unified approach to address problems on qualitatively different systems. This thesis also investigates, via two distinct avenues, the efficient numerical solution to optimal control problems of interest. Firstly, it brings together the Markov chain ap- proximation methods of Kushner-Dupuis and the triangulation of manifolds introduced by Munkres, in order to numerically solve the optimal control problem on quantum systems via a value iteration approach. Secondly, the exponential growth in computational resources required - termed the curse of dimensionality (COD) -inherent in any spatial discretization approach to solving the HJB PDE is drastically lessened via a new reduced complexity method motivated by the COD free max-plus technique recently developed by McEneaney et al. The dramatic speedup obtained by this approach enables an approximate solution to a previously intractable optimal control problem for a two qubit2 spin system evolving on a 15 dimensional Lie group.

Author dc:creator

• Sridharan, Srinivas

Statement dc:rights

• Author retains copyright

Language dc:language.iso

en_AU

Dc Identifier Other

b2569833

OAI identifier oai:identifier

oai:openresearch-repository.anu.edu.au:1885/150521