Hybrid Systems:  A Two-time-scale Approach

 
Professor George Yin

 

Department of Mathematics

Wayne State University

 

Owing to applications in signal processing, random evolution, telecommunications, financial engineering, and production planning, there has been renewed interest in understanding hybrid and large-scalesystems, where the usual dynamics and discrete events coexist. These discrete events evolve randomly and influence the underlying dynamic systems by providing a set of regimes; across the regimes, the dynamic behavior of the systems can be markedly different. In this talk, we summarize some of our recent work, in which the discrete events are formulated as a finite-state random jump process. When the state space of the jump process becomes fairly large, the computational task for obtaining optimal control policies is difficult or infeasible. Aiming at reduction of complexity, we consider an alternative approach that uses time-scale separation to reveal the hierarchical structure of the underlying systems. We first derive basic properties to gain a comprehensive understanding of the jump process; we then develop near-optimal controls of the hybrid systems.

 

Friday, September 12, 2003

3:30 – 4:30 p.m.

1500 EECS