Pre- and Post-Contact Policy Decomposition for Planar Contact Manipulation Under Uncertainty

摘要

We consider the problem of using real-time feedback from contact sensors to create closed-loop pushing actions. To do so, we formulate the problem as a partially observable Markov decision process (POMDP) with a transition model based on a physics simulator and a reward function that drives the robot towards a successful grasp. We demonstrate that it is intractable to solve the full POMDP with traditional techniques and introduce a novel decomposition of the policy into pre-and post-contact stages to reduce the computational complexity.