Welcome to the Interactive and Trustworthy Robotics Lab! Our goal is to enable embodied AI agents to reliably interact with the open world.
This includes---but is not limited to---quantifying a robot's confidence (e.g., in a requested task or in their predictive models of the world), computing safe interaction policies (e.g., that prevent nuanced safety hazards like tearing, spilling, or breaking), and aligning AI models with hard-to-specify end-user values.
Towards this, our research develops theory and algorithms that draw upon optimal control and reinforcement learning, dynamic games, uncertainty quantification, and deep learning.
Our work is applied in a variety of domains, such as lightweight robotic arms, quadrotors, quadrupedal robots, and autonomous cars.
For a detailed research overview, please see our publications page.
Congratulations to Dr. Ravi Pandya and Dr. Ran (Thomas) Tian! They are the first two PhD graduates from the Intent Lab and are off to do big things as research scientists.
New arXiv paper on a general calibrated regret metric for detecting and mitigating human-robot interaction failures; check out the applicability to anomaly detection in interactive generative planners too!
