Decision Making
We develop decision-making and planning systems that weigh a human teammate's physical and mental condition when selecting robot actions. A central goal is interpretability: robot behavior should read as legible and unobtrusive to the people working alongside it, which shapes our work on human-aware planning and decision modules.
← All research areas
Related work
-
Safe Explicable Planning Featured
Planning and policy search methods that stay both formally safe and interpretable to a human observer.
Robot behavior that is technically safe can still be confusing or unnerving to a human teammate if it doesn’t match their expectations. This line of work develops planning and policy search methods that jointly optimize for safety guarantees and explicability (behavior a human observer can readily interpret as working toward the task at hand), producing human-aware planning and decision modules suited to close collaboration.
-
Multi-Robot Task Allocation for Multitasking Robots
Assigning tasks across robot teams where individual robots can hold and interleave more than one job at a time.
Most multi-robot task allocation work assumes each robot handles one task at a time. This project relaxes that assumption, developing allocation algorithms for teams of robots that can multitask (interleaving several jobs concurrently) and studying how that changes what an efficient, human-legible allocation looks like.