A tight integration of task planning and motion planning in an execution monitoring framework

Abstract

We study integration of task planning and motion planning in a general execution monitoring framework. We show the applicability of this framework, on a modified version of Tower of Hanoi where the orientations of rings matter and rotations of rings maybe required while being moved. We represent this problem as a planning problem in the action language CCalc, and compute a shortest task plan using the reasoning system CCalc. A collision-free trajectory for the task plan (if one exists) is computed by a motion planning algorithm, using Rapidly exploring Random Trees. The monitoring agent invokes replanning upon two kinds of failures: when a continuous trajectory for a task plan cannot be computed, and when the execution of the plan fails due to external intervention. We demonstrate the effectiveness of our approach by implementing it using two robotic manipulators on a physical experimental setup.