Theory and algorithms for optimization-based control on networks

A huge amount of available data and ever-increasing computing capability enable control engineers to analyze systems of unprecedented scale and to design sophisticated controllers for advanced tasks. To benefit from these resources, large-scale control systems are often connected to networks (the internet, for example), and therefore are facing new challenges. Such challenges include computing an optimal control strategy in a timely manner, and providing security and privacy guarantees while maintaining satisfactory system performance. My research focuses on developing novel algorithms to solve these emerging problems. I will discuss inexact splitting methods (also known as inexact alternating direction methods in optimization), and show how to use them to solve distributed model predictive control (MPC) problems with limited computation and communication resources. I will further talk about novel distributed optimization algorithms with privacy guarantees on local information and discuss the trade-off between the algorithm performance and the privacy level. Finally, I will introduce a plug-and-play MPC method for interconnected linear systems, which addresses the challenge of performing network changes during closed-loop operation, while maintaining stability and constraint satisfaction.
Ye Pu is currently a postdoctoral researcher in the Hybrid Systems Lab at the University of California, Berkeley, from September 2016. She received a BS degree and an MS degree for Electrical Engineering from Shanghai Jiao Tong University, China, in 2008 and the Technical University of Berlin, Germany, in 2011, and a Ph.D degree from Swiss Federal Institute of Technology in Lausanne (EPFL), Switzerland, for her work "splitting methods for distributed optimization and control" in 2016. Her current research interests include optimization-based control (model predictive control), distributed algorithms and optimization theory with applications in power networks and multi-robot systems.