Somil Bansal
Stanford University, CA·Engineering·Early-career
Self-driving cars and autonomous drones decide what to do based on what their cameras and sensors see. Machine learning interprets that data.
- Fields
Engineering
Computer Science
- Topics
robotics
robot safety
autonomous systems
- Grants
- 1 known grant, through May 2028
What they work on
- Develop training techniques that make machine-learning perception policies robust by construction, so camera-based controllers stay safe in new situations.
- Build runtime safety monitors that detect when a perception-based controller may fail while the robot is operating.
- Create a principled method to mine closed-loop perception failures at scale and use them to improve robot safety over time.
- Evaluate these methods on a physical robot testbed and photorealistic simulators for autonomous inspection and aircraft landing tasks.
More about the research
When this perception fails, the mistake can cascade into a crash, as recent self-driving car accidents have shown. This project asks how to guarantee that a robot stays safe when its controller relies on learned perception.
Why it matters
Reliable safety guarantees for perception-driven robots could make it practical to use them where failures are costly. Examples include autonomous cars, delivery and service robots, and surveillance of critical infrastructure. Mining real perception failures could help engineers and regulators find and fix weaknesses before they cause accidents.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Sayan Mitra
University of Illinois Urbana-Champaign, IL·Computer Science
Develop refinement algorithms that synthesize perception contracts, machine-checkable bounds on how wrong an AI perception module can be, together with the operating design domains in which they hold, for pipelines that fuse camera and inertial sensor data.
Marcelo d'Amorim
North Carolina State University, NC·Computer Science
Improve the efficiency of simulation-based testing of autonomous driving systems by using risk signals from anomaly detectors during simulations to find infractions within a fixed time budget.
Jaime Fernández Fisac
Princeton University, NJ·Engineering·Early-career
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.
The school
Stanford University: 87 labs on PI Pages, $1.6B a year on research, 739 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.