Past Event: Oden Institute Seminar

Control, architecture design, and learning in dynamical networks with multiplicative noise

Tyler Summers, UT Dallas

Abstract

Emerging highly distributed networked dynamical systems, such as critical infrastructure for power, water, and transportation, are increasingly being instrumented with new sensing, actuation, and communication technologies. This is presenting many challenges and opportunities for more sophisticated control, optimization, and learning algorithms and architectures to enhance performance and robustness. In this talk I will present our recent work on control, architecture design, and learning in dynamical networks with multiplicative noise. Multiplicative noise models have a long history in control theory but are re-emerging as compelling uncertainty representations in networks and systems with data-driven learning and adaptation that can improve robustness of controllers. I will show that although policy optimization algorithms require optimization of a non-convex cost function, the multiplicative noise linear quadratic regulator cost has a special property called gradient domination, which is exploited to prove global convergence to the globally optimal policy. This framework is then utilized for control architecture design using sparsity-promoting regularization and various learning-based control algorithms. Bio: Tyler Summers is an Assistant Professor in the Departments of Mechanical and Electrical Engineering at UT Dallas, where he directs the Control, Optimization, and Networks Laboratory. Prior to joining UT Dallas, he was an ETH Postdoctoral Fellow in the Automatic Control Laboratory at ETH Zürich from 2011 to 2015. He received a PhD degree in Aerospace Engineering with emphasis on feedback control theory at the University of Texas at Austin in 2010. He was a Fulbright Postgraduate Scholar at the Australian National University in Canberra, Australia in 2007-2008. He is the recipient of a CISE Research Initiation Initiative (CRII) award from the National Science Foundation in 2016 and a Young Investigator Program award from the Army Research Office in 2017. His research interests are in distributed feedback control, optimization, and learning in complex dynamical networks, with applications to distributed robotic networks and electric power networks.