Efficient planner using information of hierarchy and the Monte-Carlo tree search

Abstract

We propose a hierarchical planning algorithm that efficiently finds the optimal plan for the given planning problem. A key challenge to the hierarchical planning approach is that a higher-level plan may fail to properly guide the planning at the lower level. To handle this we view this problem as the balancing problem between exploration and exploitation and use the Monte-Carlo tree search algorithm to efficiently focus on seemingly promising high-level plans. We also present a theoretical analysis on the optimality of our algorithm and show that under certain conditions, our algorithm is guaranteed to find the optimal plan. We showed that our method outperforms existing algorithms in the continuous path planning domain.