Manipulation Planning with Tight Geometric Constraints

Hierarchical task and motion planning guided by a reachability tree.

← Back to Physical AI

Hierarchical task and motion planning guided by a reachability tree.

Overview

Manipulation tasks with tight geometric constraints are difficult for sampling based methods because many sampled actions are infeasible. This project introduces a hierarchical planning algorithm built around a reachability tree to focus computation on promising task and motion choices.

Results from low level motion planning are returned to the high level planner. This experience improves later decisions, reduces unproductive sampling, and avoids unnecessary growth of the symbolic planning problem.

Video

Publication

K. Kim, D. Park, and M. J. Kim, “A Reachability Tree-Based Algorithm for Robot Task and Motion Planning,” IEEE ICRA, 2023. arXiv · IEEE