•   When: Thursday, March 05, 2020 from 11:00 AM to 12:00 PM
•   Speakers: Gregory Stein
•   Location: Engineering Building 4201
•   Export to iCal

Abstract:

Prototype self-driving vehicles populate our roads, providing mobility to those who need it, and home healthcare robots promise to provide independence to those who lack the ability to care for themselves. Yet to realize their full potential, robots must become more capable of operating in a world not built for them: to plan intelligently in environments designed for and populated by people. Planning in a dynamic and complex world means that, to act intelligently, embodied agents must plan under uncertainty and reason about the implications of their actions many steps into the future, a process that typically requires enormous computational effort.

My work, at the intersection of embodied intelligence and machine learning, is centered around developing new representations that enable embodied agents, like robots, to better understand the impact of their actions, so that they may plan quickly and intelligently in a dynamic and uncertain world. I have created tools that leverage machine learning to robustly build these representations from imperfect sensor observations and to estimate interpretable properties about the world that allow the agent to efficiently make predictions about the future. In this talk, I focus on my recent efforts to advance autonomous navigation in unknown environments, and discuss how the decision-making paradigm I have created allows for high-level navigation that shows reliable performance in both simulation and on real-world hardware.

Bio: Gregory J. Stein is a Ph.D. candidate at the Computer Science and Artificial Intelligence Lab in the MIT Department of Electrical Engineering and Computer Science. He previously received an S.M. from the MIT Department of Electrical Engineering and Computer Science and graduated summa cum laude from the Cornell University Department of Applied and Engineering Physics. His work was a finalist for Best Paper at the 2018 Conference on Robot Learning, at which he was additionally awarded Best Oral Presentation. He was also awarded the Trevor R. Cuykendall Award for Outstanding Teaching Assistantship from the Cornell University Department of Applied and Engineering Physics.