Computing Over Unreliable Communication Networks

ECE Seminar: Computing Over Unreliable Communication Networks

Starts at: October 16, 2014 4:30 PM

Ends at: 6:00 PM

Location: Scaife Hall 125

Speaker: Nicola Elia

Affiliation: Iowa State University

Refreshments provided: Yes


In this talk, we take the unifying view of systems interacting overcommunication networks as distributed computing systems and propose tostudy them as networked control systems. Since averaging is a centraloperation to much science and engineering, we first study the problem ofdistributed averaging over unreliable networks. We show that a popularand well-behaved algorithm can instead generate a collective globalcomplex behavior when the inter-agent communication happens overunreliable links. We characterize this behavior, common to many naturaland human-made interconnected systems, as a collective hyper-jumpdiffusion process and a Levy flight process in a special case. To mitigate the effects of the unreliable information exchange, we propose a newdistributed averaging algorithm resilient to noise and intermittentcommunication. The algorithm and the control perspective are the basis forthe development of new distributed optimization systems that we cananalyze and design as networked control systems. The approach applies tomulti-agent cooperative applications and opens up several directions ofresearch.
Nicola Elia is a Professor of Electrical and Computer Engineering at Iowa StateUniversity. He received the Laurea degree in Electrical Engineering from Politecnico diTorino in 1987, and the Ph.D. degree in Electrical Engineering and Computer Sciencefrom Massachusetts Institute of Technology in 1996. He worked at the Fiat ResearchCenter from 1987 to 1990. He was Postdoctoral Associate at the Laboratory forInformation and Decision Systems at MIT from1996 to 1999. He has received the NSFCAREER Award in 2001. His research interests include networked control systems,communication systems with access to feedback, complex systems, distributedoptimization and control.