Abstract:

Existing computational tools for the analysis of dynamical systems do not scale well to large-scale networked systems. In this talk we present a compositional method to certify performance and safety properties of interconnected systems. Using the dissipativity properties of each individual subsystem we formulate the certification problem as a large-scale optimization problem. This problem searches for the dissipativity properties of the subsystems that are most relevant in terms of certifying the specified performance or safety property. Distributed optimization techniques, specifically the alternating direction method of multipliers(ADMM), is employed to decompose and efficiently solve this problem.

Biography:

Chris Meissen is a Ph.D. candidate at the University of California, Berkeley in the Berkeley Center for Control and Identification. He is under the supervision of Dr. Andrew Packard and Dr. Murat Arcak. He received his B.S. from Kansas State University in 2007 and his M.S. from Colorado State University in 2009, both in Mechanical Engineering. His research interests include dynamical system analysis, robust and nonlinear control theory, and large-scale optimization, especially semidefinite and sum-of-squares programming.