Decision and Control Lecture Series
Decision and Control Laboratory, Coordinated Science Laboratory
Designing Consensus Protocols
Department of Industrial & Enterprise Systems Engineering
University of Illinois
Wednesday, September 4, 2013
3:00 a.m. to 4:00 p.m.
We discuss the design of fast protocols for the so-called average consensus problem: computing the average of numbers stored throughout the network in a way that is fully distributed, requires no centralized supervision, robust to unpredictable node and link failures, and works with only unidirectional communication links. Consensus protocols can be applied to solve a variety of multi-agent control problems, and we will describe some recent work showing how to use them to solve some cooperative estimation and optimization problems when communication and sensing among agents is time-varying and directed. We then turn to the main theme of the talk, which is designing consensus protocols which work fast, and describe a consensus protocol with the best known convergence time to date.
Alex Olshevsky received his Ph.D. in Electrical Engineering in Computer Science from MIT in 2010. After spending two years as a postdoc in the Mechanical and Aerospace department at Princeton, he joined the ISE department at UIUC in 2012. His research interests lie in control theory and optimization, especially applied to distributed and multi-agent systems. He is a recipient of the ICS Prize from INFORMS in 2012 as well as a paper award from SIAM in 2011.