H.B. Keller Colloquium

Abstract: Recent radical evolution in distributed sensing, computation, communication, and actuation has fostered the emergence of cyber-physical network systems. Examples cut across a broad spectrum of engineering and societal fields such as power grids, swarm robotics, air/ground transportation systems, green buildings, and other societal networks. Regardless of the specific application, one central goal is to shape the network collective behavior through the design of admissible local decision-making algorithms. This is nontrivial especially due to the challenges placed by the local connectivity, model and environment uncertainties, imperfect communication, and the complex intertwined physics and human interactions.  In this talk, I will present our recent progress in formally advancing the systematic design of distributed coordination in network systems via harnessing special properties of the underlying problems and systems. In particular, we will present three examples and discuss three type of properties, i) how to use network structure properties to develop scalable reinforcement learning algorithms, ii) how to use system physical dynamics to develop real-time feedback optimization algorithms, and iii) how to use algorithmic properties to develop communication-reduced distributed algorithms.

Acknowledgement: The talk is based on different pieces of work with many collaborators. First authors include Guannan Qu, Sindri Magnusson, Yingying Li, Xin Chen, Yujie Tang.