Event Location: GHC 8102

Abstract: We want our robots to work in human environments. However, human environments are cluttered and clutter creates problems for robotic manipulation. We propose to develop a framework for manipulation in highly cluttered environments. We propose to rearrange the environment using prehensile and non-prehensile actions. We present our preliminary work, a framework in which a high-level rearrangement planner utilizes a low-level library of open-loop actions. We propose to extend the capabilities of our existing framework by: (i) Utilizing sensor feedback for the control of low-level actions; (ii) Utilizing feedback at the high-level of rearrangement planning; (iii) Reformulating the planner so that it reasons about the spaces occluded by clutter and plans to move objects to see the space behind them; (iv) Designing new non-prehensile actions that can move multiple objects simultaneously.

Committee:Siddhartha Srinivasa, Chair

Chris Atkeson

Matthew T. Mason

Tomas Lozano-Perez, Massachusetts Institute of Technology