Memory-based hierarchical task and motion planning

It has been shown in recent robot planning control research that planning with simplified domain models is efficient and can be robust by detecting execution failures and replanning online. In this work, we alter the traditional HTN planning by interleaving with execution and extend it with a geometric planner for mobile manipulation tasks. We also build a memory-base which is comprised of several sub-memories. It can memory and manage mass patter-specific records, and supply data with the most efficient patter to the planner and various actuators. At the last, we present preliminary results on a mobile robot with manipulator and shows some advantage of IPAE when robot encounter uncertainty and need replanning, compared with the traditional HTN method.