Jaime Fernández Fisac
Princeton University, NJ·Engineering·Early-career
Robots are moving into homes, streets and skies. Robotic caregivers, delivery drones and self-driving cars all have to operate around people.
- Fields
Engineering
Computer Science
- Topics
robot safety
safe autonomy
dynamic game theory
- Grants
- 1 known grant, through Jul 2029
What they work on
- Develop a framework based on dynamic game theory in which a robot guarantees safety against the worst case within an Operational Design Domain (ODD), the conditions designers or regulators say it must handle, while modeling how agents gain information over time.
- Use deep adversarial reinforcement learning, training against a simulated opponent, to find approximate safe strategies for legged walking robots and dense urban driving under shoves, changing terrain, poor visibility and misleading human behavior.
- Turn these untrusted learned strategies into verified safety filters using real-time robust planning and control, so that limited computing power can only make a robot more cautious, never unsafe.
- Validate the safety filters on legged robot hardware, autonomous driving datasets and simulators, and in-lab user studies of human–robot interaction.
More about the research
Yet they lack the kind of safety guarantees we expect from bridges, elevators and airplanes. Older safety methods that worked in factories struggle in the messy outside world. Machine learning, which relies on large amounts of data, can be caught off guard by rare but dangerous events. This project aims to let robots continually prove that their next action is safe against any possible event within clearly defined conditions. For example, a self-driving car should avoid a collision even if a nearby cyclist suddenly falls into its path.
Why it matters
Self-driving cars and mobile robots will only be trusted around people if their safety can be clearly shown. Safety filters with checkable guarantees could make these robots more reliable. They could also give regulators a clear basis for certification and help separate unavoidable accidents from true safety failures. The insights may extend to the safety of other AI systems.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Sandhya Saisubramanian
Oregon State University, OR·Computer Science·Early-career
Develop model cognizance methods that let autonomous agents proactively identify and repair limitations in their symbolic decision-making models through limited user interaction.
Yorie Nakahira
Carnegie Mellon University, PA·Engineering·Early-career
Quantify long-term risks in autonomous control systems when key variables are hidden and data on rare risk events are limited.
Somil Bansal
Stanford University, CA·Engineering·Early-career
Develop training techniques that make machine-learning perception policies robust by construction, so camera-based controllers stay safe in new situations.
The school
Princeton University: 26 labs on PI Pages, $559M a year on research, 403 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.