 |
Professor of Electrical EngineeringSchool of Electrical Engineering and Computer Science
|
Distributed Algorithms for Networked Control Systems
Abstract: In many applications, a network of autonomous agents holds eminent promises to achieve a desired level of performance, capability, robustness, and efficiency beyond what a single agent can achieve. However, to be advantageous, multiple agents have to work in an coordinated and synchronized manner. This talk considers a networked control system involving a set of autonomous agents (or nodes) and seeks a distributed algorithm to steer the agents so that a network-wide objective is achieved. The first focus of this talk is to discuss some newly proposed fast converging distributed algorithms. We will start with the well-known average consensus problem and provide a new distributed algorithm which converges rapidly in comparison with the commonly used Laplacian matrix based algorithm. We then move on to several distributed estimation problems, including weighted least squares (WLS) estimation, maximum likelihood estimation, and linear equations, and provide similarly fast distributed algorithms. These algorithms all have the common features that they converge in a finite number of iterations when the network graph is acyclic (without loops) and that they converge asymptotically for a general graph under some mild conditions. The second focus of this talk is to investigate the formation control for a networked multi-agent systems. We will present some new distributed control solutions for 2-dimensional and higher dimensional formation.