Upcoming Event: Center for Autonomy Seminar

High Performance Control of A Large Network of Inexpensive Agents

Dr. Faryar Jabbari, Professor of Mechanical and Aerospace Engineering

Abstract

There have been tremendous advancements in high-performance robots, drones, and other autonomous or semi-autonomous systems. At the same time, interest is growing in use of large networks of inexpensive agents for a variety of tasks. This often creates challenges due to limitations on power (actuation), bandwidth (communication frequency), as well as communication delays, heterogeneity of agents, or configuration reliability (loss or addition of some agents during operation). For a single agent, strong results exist for most of these challenges. Of course, all inevitably result in some level of conservatism e.g., degradation of performance guarantees. It is possible to extend these techniques to multi-agent systems, while preventing the size of the underlying problem from growing with the number of agents.  This, however, leads to an additional source of conservatism due to the structure of the network interaction, as captured by the Adjacency or Laplacian matrices.  The nature of the connectivity of a network may pose an inevitable loss of performance, compared to that of a single agent. There often is additional conservatism due to the selection of weights in the Laplacian.  For a known network configuration, batch processes can optimize these weights, but such optimized weights may become ineffective when the network undergo changes due to loss or gain of agents or reconfiguration amongst the agents. This talk presents an approach for agents to, autonomously and decentrally, optimize their network weights for improved performance, in real time and with low computational sophistication. The proposed solution is a mixture of known -- in original form or somewhat modified - techniques: max consensus, augmented Lagrangian, power iteration, etc., enabling autonomous operation of a network of inexpensive networked agents, under possibly hazardous conditions, with optimized performance.

Biography

Faryar Jabbari is on the faculty of the Mechanical and Aerospace Engineering Department of University of California, Irvine. His research interests are in control and its applications (structural or energy systems).  He has been active in IFAC and IEEE, serving in a variety of roles, such as program or general chair for ACC and CDC conferences.  He is currently serving as the Interim Dean of the Henry Samueli School of Engineering of the University of California, Irvine.