A sampling-based planner for underground pipeline inspection with unknown narrow passage

Abstract

Reliable path planning is critical for pipeline inspection robots operating in constrained industrial environments. Sampling-based methods have the advantage of high computational efficiency and are widely deployed in robot path planning. However, they face two challenges in underground pipeline inspection: poor adaptability in complex narrow passage scenarios and how to autonomously plan the path fitting pipeline layout. To address the above challenges, we propose a path planner RGP-RRT* (Ring Guide Pipeline-RRT*). First, we propose the ring guide mechanism that dynamically adjusts sample density near obstacle boundary based on the status of node extension. It enhances the adaptability to different narrow passage scenarios and provides orientation guidance for the planner. Second, we propose a pipeline-biased sampling method, which samples the area near the underground pipeline based on the sensor detection radius. This method improves search efficiency while ensuring that the planned path maintains sensor coverage to the underground pipeline. In addition, a pipeline-aligned post-processing method is proposed, which autonomously generates the path fitting the pipeline layout while considering obstacle avoidance. Simulations and experiments show that our planner improves time cost, success rate, and inspection performance.