The author provides a comprehensive study of averaging methods for singularly perturbed control systems, including averaging over fast control-state trajectories and over the occupational measures generated by such trajectories. Averaging of Singularly Perturbed Control Systems and LP Approach to Optimal Control establishes rigorous connections between optimal control problems with time-discounting or time-averaging criteria and infinite-dimensional LP problems, proves strong duality for the latter and uses this duality to analyze and construct optimal and near-optimal controls. It also develops LP-based techniques for the analysis and construction of asymptotically near-optimal controls for singularly perturbed systems. These techniques are illustrated by numerical examples.
Vladimir Gaitsgory is an Honorary Professor at Macquarie University, Australia. His research interests include control, optimization, dynamical systems, game theory, and their applications. His main contributions are to the development of averaging methods for singularly perturbed control systems and LP-based approaches to optimal control.