Modeling, Analysis, and Design of Influence in Multi-Agent Systems

Abstract: Systems of intelligent agents interacting according to their own policies may yield behavior that is contrary to the social good of the community. To achieve regulatory control objectives that change the group’s equilibrium behavior, an intelligent central planner (CP) must understand the learning mechanisms at the individual level, characterize how global intervention disrupts these learned action processes, and choose control policies that induce the desired change. This framework describes phenomena such as social media, financial networks, and cyber-physical systems like power grids. In this talk I model this influence structure as a Markov decision process (MDP) with the CP as the controller. I characterize the CP’s capabilities for a given scenario by analyzing the reachability of control objectives and finding policies that attain reachable objectives. I discuss how to implement cluster-based control policies, from how to efficiently compute near-optimal clustered policies to using properties of submodular optimization to assign agents to clusters. Next, I consider the problem of model-free policy design that is robust to agent dropout. First, game theoretic techniques measure the potential impact of each agent on the CP’s value function, and the desired robustness criterion is embedded into the MDP.  The post-dropout MDP can be evaluated with high probability via policy importance sampling, and safe policy search routines find desirable robust policies while maintaining a baseline value. Future work is motivated that would enable more sophisticated control techniques to handle systems at scale and with greater complexity.

Bio: Bruno Sinopoli is the Das Family Distinguished Professor at Washington University in St. Louis, where he is also the founding director of the center for Trustworthy AI in Cyber-Physical Systems and chair of the Electrical and Systems Engineering Department. He received the Dr. Eng. degree from the University of Padova in 1998 and his M.S. and Ph.D. in Electrical Engineering from the University of California at Berkeley, in 2003 and 2005 respectively. After a postdoctoral position at Stanford University, Dr. Sinopoli was member of the faculty at Carnegie Mellon University from 2007 to 2019, where he was a professor in the Department of Electrical and Computer Engineering with courtesy appointments in Mechanical Engineering and in the Robotics Institute and co-director of the Smart Infrastructure Institute. His research interests include modeling, analysis and design of Resilient Cyber-Physical Systems with applications to Smart Interdependent Infrastructures Systems, such as Energy and Transportation, Internet of Things and control of computing systems. More recently, he has been working on understanding connections between Machine and human learning and influence mechanisms in multi agent systems.