Yorie Nakahira
Carnegie Mellon University, PA·Engineering·Early-career
Self-driving cars, medical devices, assistive robots and factory machines increasingly make decisions while working alongside people in unpredictable settings. Keeping these autonomous systems safe is hard.
- Fields
Engineering
Computer Science
Mathematics & Statistics
- Topics
control theory
safe control
autonomous systems
- Grants
- 1 known grant, through Jun 2030
What they work on
- Quantify long-term risks in autonomous control systems when key variables are hidden and data on rare risk events are limited.
- Develop real-time control methods based on probabilistic invariance, keeping a system within safe bounds with high probability, to guard against latent and human-perceived risks over long horizons.
- Infer human-perceived risk and safe actions from sparse human feedback so that controllers can reduce the risk people actually feel.
- Design control strategies for repeated, changing human–machine interactions that avoid encouraging aggressive or exploitative human adaptation and instead promote desirable behavior.
More about the research
Common risk methods assume the system's full state and dynamics are known and that plenty of examples of rare dangerous events exist, which is often untrue in practice. Systems must also account for risks that people perceive, not just measurable ones. In addition, people adapt over repeated encounters, so a controller that ignores this may unintentionally encourage aggressive or exploitative human behavior.
Why it matters
Autonomous systems in transportation, healthcare, assistance and manufacturing must stay safe around people despite hidden variables, scarce data on rare failures, and changing human behavior. Control methods that give long-term safety guarantees and account for how humans perceive risk and adapt could make these systems safer and more trustworthy. They could also help avoid designs that unintentionally invite risky or exploitative human behavior.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
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.
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.
Nikolai Matni
University of Pennsylvania, PA·Engineering
Develop safety-rich data augmentation and domain randomization techniques for training learning-enabled systems, and determine which types of data must be collected to guarantee end-to-end safety.
The school
Carnegie Mellon University: 28 labs on PI Pages, $505M a year on research, 346 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.