Hierarchically Accelerated Coverage Path Planning for Redundant Manipulators

TL;DR

This work addresses coverage path planning for redundant robot manipulators by exploiting task tolerances to minimize joint-space costs. It reframes the problem as a Generalized Traveling Salesman Problem (GTSP) and introduces a hierarchical method, H-Joint-GTSP, that identifies exemplar poses to constrain IK sampling and solves smaller GTSPs to derive a joint-space path. Compared with Cart-TSP-IKLink and Joint-GTSP baselines, the proposed approach achieves higher-quality motions with shorter computation times across four simulated benchmarks and a real-robot demonstration, while reducing arm reconfigurations. By leveraging redundancy and a two-level planning strategy, the method improves scalability and practical applicability for tasks like polishing, wiping, and surface inspection. An open-source implementation accompanies the work, enabling broader adoption and further development.

Abstract

Many robotic applications, such as sanding, polishing, wiping and sensor scanning, require a manipulator to dexterously cover a surface using its end-effector. In this paper, we provide an efficient and effective coverage path planning approach that leverages a manipulator's redundancy and task tolerances to minimize costs in joint space. We formulate the problem as a Generalized Traveling Salesman Problem and hierarchically streamline the graph size. Our strategy is to identify guide paths that roughly cover the surface and accelerate the computation by solving a sequence of smaller problems. We demonstrate the effectiveness of our method through a simulation experiment and an illustrative demonstration using a physical robot.

Figures (3)

• Figure 1: An illustration of the three approaches described in this paper and evaluated in our experiment.
• Figure 2: Our experiment involves four benchmark applications. For each, we provide visualizations of the robot performing the task, task tolerances, and how the surface is covered using different approaches. We optimize for the number of reconfigurations (breakpoints on the paths) and joint movements; motions with shorter joint movements tend to result in complex Cartesian space paths. The robot visualizations and green traces were generated using Motion Comparatorwang2024motion.
• Figure 3: To demonstrate the scalability of the approaches, we increased the number of end-effector targets sampled on the surface in the wok polishing task. The results show that the proposed method, H-Joint-GTSP, has superior scalability compared to both Cart-TSP-IKLink and H-Joint-GTSP because it consistently has faster performance and produces higher-quality motion, even when applied to increased sampling density of the input.