The Google award comes for his work in “machine learning and data mining,” one of 16 research topics Google has targeted for support. All awardees hold faculty positions and have received their Ph.D.s within the previous seven years.

Fernández Fisac’s research draws heavily on applied math, cognitive science and AI to build systems that can reason about their own safety in dynamic situations. That’s especially important with the rapid growth of autonomous systems deployed in human-centric environments. For example, with millions of self-driving cars on the road, one-in-a-million scenarios would become daily accidents, many of them catastrophic. Fernández Fisac, assistant professor of electrical and computer engineering, sees guarding against those relatively low-frequency, high-risk encounters as one of the essential problems facing modern robotics.

The award-winning project involves robots learning robust safety strategies in a game-like simulation, similar to the way other systems learned to beat humans in complex games such as Go. This simulation involves ratcheting up the elements of Murphy’s Law so that the robots face a full range of unlikely, but possible, dilemmas. The technique at the center of that training, called adversarial reinforcement learning, gives the robots tools to reason about safety in uncertain physical environments, where they must stay one step ahead of possible no-win scenarios to make informed, safety-critical decisions.

Fernández Fisac joined Princeton in 2020 after a year at Waymo, formerly the Google self-driving car project. Prior to that he earned his Ph.D. at the University of California-Berkeley, where he received the Leon O. Chua Award for excellence in nonlinear science. Last year, he developed a course called Intelligent Robotic Systems, aimed at teaching undergraduates to use industry-standard tools for the design and deployment of advanced robotic systems.