CDS/EE - Estimation and Control in Networked Systems

Vijay Gupta
Ph.D. candidate in the Electrical Engineering Department
California Institute of Technology.

Abstract: Numerous applications of systems involving multiple components - sensors, actuators and controllers - working in a cooperative fashion are being envisaged today. Although we have a rich theory to design single component systems, designing networked multi-agent systems poses new challenges and requires new design principles. A fundamental challenge in multi-agent systems is ensuring cooperation among components that have access to different information sets. To cooperate, components need to exchange information over communication channels which introduce various insidious effects such as data loss, data corruption, random delays and so on and severely degrade performance.

In the first part of my talk, I will consider a problem that arises frequently in multi-agent systems - estimating and controlling a dynamic process over a communication network. The performance degradation due to channels is showcased starkly in such networked control systems. I will discuss one approach to counter channel effects, which is the use of computational resources at the component level to preprocess the information that needs to be communicated. In particular, I will identify simple yet optimal preprocessing algorithms to deal with stochastic data loss by communication channels. These algorithms, in a sense, correspond to optimal coding strategies for the purpose of estimation and control.

In a short second part of the talk, I will touch upon some other problems that arise in the control of networked systems.

Bio: Vijay Gupta is a Ph.D. candidate in the Electrical Engineering department at the California Institute of Technology. He received his B.Tech degree at the Indian Institute of Technology, Delhi in 2001 and his M.S. degree at Caltech in 2002, both in Electrical Engineering. His research interests include networked control systems and sensor networks, distributed estimation and detection, usage-based value of information and the interaction of communication, computation and con