ECE 851

Stochastic Estimation and Control Systems

The Ohio State University

Winter Quarter 2008

Prerequisites: ECE 750, and ECE 650 or ECE 805

Instructor: J. B. Cruz, Jr.

Call Number: 10550-3

MWF 11:30-12:18 PM Caldwell Laboratory, Room 0137

CATALOG DESCRIPTION

851 * Stochastic Estimation and Control Systems G 3
Synthesis of systems, both linear and nonlinear, with random inputs; advanced topics.

Course Goals

The first objective of this course is to provide a graduate level development of optimization of stochastic systems, principally stochastic dynamic programming, for both perfect state and imperfect state information. In the case of imperfect state information, state estimation is part of the control optimization. The second objective of the course is to provide an introduction to optimization of multi-agent and multi-team decision making in stochastic systems.

Course Outline

I. Dynamic Programming for Stochastic Dynamic Discrete-time Control Problems

A. Discrete-time nonlinear processes. Formulation of optimization by dynamic programming.

B. Derivation of optimal control for linear systems with noisy inputs, noisy observations, possibly random parameters in the plant, quadratic cost functionals, and Gaussian noise. Certainty Equivalence and the Separation Principle: deterministic LQ controller and Kalman filter.

C. Minimum variance control using ARMAX models.

II. Approximations of Cost-to-Go, Suboptimal Control (rollout algorithms, model predictive control, receding horizon control)

III. Cooperative, non-cooperative, and leader-follower strategies for stochastic multi-agent, multi-team dynamic systems.

References: Readings from several books and papers.