Research Areas

Adversarial Multi-Agent Systems

We want to enable robots to operate safely and optimally in the presence of adverserial agents. Our toolbox includes game theory, reinforcement learning, and optimal control. We also want to address safety assurances in learning-based control systems.

As a demonstrative scenario, part of our research deploys systems on autonomous race cars that are capable of reaching 200 MPH, and battle against other opposing agents. We participate in the Indy Autonomous Challenge (IAC) and are the top U.S. team.

This research project is in collaboration with UC Berkeley EECS, UC San Diego, and University of Hawaii through AI Racing Tech and DARPA Assured Neuro Symbolic Learning and Reasoning (ANSR) program.