Search:

Helping Human-robot Teams Stay Coordinated

Colloquium

Speaker(s)
Dr. Jim Boerkoel
Date
Thursday, April 11, 2013
Time
4:15 PM – 5:30 PM
Location
Galileo Pryne

An emerging trend in robotics is the development of inherently safe, mobile robots that are designed to assist humans across a wide variety of tasks. Humans are good at intuitively perceiving and quickly adapting to the intentions and conventions of their human teammates with little or no explicit communications. Robots, on the other hand, face the challenge of optimizing and adapting their schedules to, what to them appears to be, their inherently unreliable, non-communicative, and irrational human teammates. A primary goal of my work is to equip robots with methods to quickly and robustly develop plans that (1) adapt to established conventions to successfully negotiate human-oriented environments and (2) augment the workflow of human workers to increase overall productivity, safety, and quality. In this talk I discuss a new, distributed robotic scheduling approach that builds on classical shortest-path algorithms to enable a robot to coordinate its activities with its human teammates while providing them flexibility and autonomy.

Biography

Jim Boerkoel is a Postdoctoral Associate with the Interactive Robotics Group in the Computer Science and Artificial Intelligence Laboratory at MIT. He is the lead researcher on a project that investigates developing and deploying mobile robot assistants in the final assembly of automotive manufacturing processes to increase the productivity, safety, and health of human workers. More generally, he is interested in using ideas from AI and robotics to improve the coordination of teams of multiple agents, which can include human, virtual, and embodied agents. His goal is to develop techniques that augment humans’ own cognitive and physical abilities to create integrated human-robot teams that are more capable than their individual counterparts. Prior to joining the IRG, Jim completed his PhD work at the University of Michigan, developing distributed approaches for constraint-based, multiagent scheduling.