Decentralized stochastic control: Information structures, optimal solutions and approximations

Speaker: Serdar Yuksel, Dept. of Mathematics and Statistics, Queen's University, Canada

This tutorial talk is concerned with stochastic dynamic decentralized control (or team) problems and their optimal solutions. A review on information structures and basics of decentralized stochastic control will be presented. Strategic measures for team problems will be introduced; these are probability measures induced on the space of measurement and action sequences by admissible decentralized control policies. Conditions ensuring the compactness of sets of strategic measures will be established. These will lead to existence results for optimal solutions for both static and dynamic teams. Properties such as convexity of the sets of such measures will be studied; it will be shown that measures induced by deterministic policies form the extreme points of a properly expanded set of strategic measures, thus establishing the optimality of such policies. Characterizations for convexity of problems which include teams with a non-classical information structure will be presented. Finally, through an approximation of the sets of strategic measures by those induced with quantization of measurement and action spaces, asymptotic optimality of finite model representations for a large class of dynamic team problems will be established. These lead to asymptotic optimality of quantized control policies. The celebrated counterexample of Witsenhausen will be discussed throughout the talk to illustrate the salient aspects of information structures in decentralized control, and demonstrate the existence, convexity, and approximation results. (Part of this talk is based on joint work with Naci Saldi and Tamas Linder).
Category Seminar
Location: Room ES52, EDIT Building, Hörsalsvägen 9, Campus Johanneberg
Starts: 26 April, 2017, 11:00
Ends: 26 April, 2017, 12:00

Published: Fri 07 Apr 2017. Modified: Tue 11 Apr 2017