Recording: Link

Abstract: Machine Learning is being increasingly deployed in safety-critical domains, including autonomous driving, healthcare, and national security. With these applications comes a responsibility for researchers to design systems that allow for effective human supervision when possible as well as safe, autonomous operation when necessary. In this talk, I will share recent contributions across Parasuraman’s Levels of Autonomy for robot and machine learning in safety-critical settings. First, I consider the problem of adaptive robot control requiring learning at scales faster than human reaction time, i.e., an autopilot adapting to in-flight damage. Here, we develop a probabilistically-safe method to actively learn system dynamics for model-predictive control. Next, I consider human supervisory control, developing human factors design principles that enable robots to engender appropriate trust and reliance from human operators. Finally, I present an interpretable, reinforcement learning technique that extracts an exact decision tree from a neural network. This interpretable policy can equip humans with better decision-making strategies when manual control is most appropriate.

Bio: Dr. Matthew Gombolay is the Anne and Alan Taetle Assistant Professor of Interactive Computing at the Georgia Institute of Technology. He received a B.S. in Mechanical Engineering from the Johns Hopkins University in 2011, a S.M. in Aeronautics and Astronautics from MIT in 2013, and a Ph.D. in Autonomous Systems from MIT in 2017. Gombolay’s research interests span robotics, AI/ML, human-robot interaction, and operations research. Between defending his dissertation and joining the faculty at Georgia Tech, Dr. Gombolay served as a technical staff member at MIT’s Lincoln Laboratory transitioning his research to the U.S. Navy, earning him an R&D 100 Award. His publication record includes a best paper award from American Institute for Aeronautics and Astronautics, and he was selected as a DARPA Riser in 2018. He was also awarded a NASA Early Career Fellowship for his work increasing science autonomy in space.