Path-Follower: A High-Precision Sampling-Based Motion Planner for Inspection Robot

摘要

Mobile inspection robots operate in narrow industrial sites have large mass and size and are usually not allowed to perform autonomous global path planning or autonomous obstacle avoidance. Instead, the robots precisely follow pre-planned global paths and stop to wait when encountering obstacles. The common working mode of inspection robots is to stop at the inspection points on the global path, collect images or point cloud data, and sometimes need to follow the global path to pass through or enter cramped spaces. To ensure the inspection effect and the passing ability, the accuracy of following the global path and reaching the target pose is very important for the motion planning of the inspection robot. This paper presents a highprecision sampling-based motion planner for metro inspection robot, named Path-Follower. Different from the traditional dynamic window approach (DWA) algorithm, Path-Follower has a special velocity sampling limit near the target pose, a local trajectory reckoning process that considers actual velocity and kinematic constraints, and a trajectory evaluation function designed for path following and target pose reaching. Path-Follower can provide the capability of accurate motion planning without modeling dynamics for mobile inspection robots. The simulation results based on Gazebo in Robot Operating System (ROS) compared with DWA algorithm built in ROS demonstrate the effectiveness of the proposed algorithm. In path following, Path-Follower has a lower absolute trajectory error of 0.011m. In target pose reaching, Path-Follower has a lower absolute end-to-end error of 0.008m.