Combining Symbolic and Motion Planners for Rearranging Tasks in Daily-life Environments

Abstract

Humans do rearranging tasks every day. These tasks are time-consuming and tedious. Rearranging tasks are challenging because there are many problems to solve, such as identifying the items, manipulating the items, and finding a strategy to rearrange that can be completed with the robot. In this work, we focus on how to execute rearranging tasks and avoid errors in the motion planner such as not finding a valid solution. We propose to combine the symbolic planner with the motion planner using a feasibility database to confirm if the instructions received by the symbolic planner are valid or not. The final poses of the robot end effector are previously validated with the motion planner, so we avoid sending the robot to invalid poses. The proposed method was tested in a simulation environment doing a sandwich rearranging task in a convenience store setup.