Systematic Design of Decentralized Algorithms for Consensus Optimization
Friday, August 16, 2019
10:30AM – 12:30PM
Decentralized optimization algorithms are widely used in the control of networked cyber-physical systems such as the power grid, transportation networks, and multi-robot teams. In a decentralized algorithm, the nodes (agents) collectively solve an optimization problem by solving part of the problem locally and exchanging messages over a communication network. In this talk, I will present a systematic procedure for designing decentralized optimization algorithms for a special class of problems in which the objective function is a sum of local objective functions. Specifically, I will show that a decentralized optimization algorithm can be synthesized by combining an existing base optimization algorithm (e.g., gradient descent) and a consensus tracking algorithm. A major benefit of this procedure is that one can separately choose the base optimization algorithm to accommodate different types of objective functions and the consensus tracking algorithm to accommodate different types of communication networks. In addition, parameters used in the synthesized algorithm can be selected in an automated manner by numerically computing a certificate of convergence using tools from robust control theory.
Shuo Han is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Illinois at Chicago (UIC). Previously, he was a postdoctoral researcher in the Department of Electrical and Systems Engineering at the University of Pennsylvania. He received his B.E. and M.E. in Electronic Engineering from Tsinghua University in 2003 and 2006, and his Ph.D. in Electrical Engineering from the California Institute of Technology in 2014. His research interests lie broadly in the areas of optimization and control theory with applications in large-scale interconnected cyber-physical systems such as transportation networks and the power grid.
Hosted by Ufuk Topcu