An Obstacle Avoidance Method Based on Advanced Rapidly-exploring Random Tree for Autonomous Navigation

Chengzhen Meng, Hongjun Dai

Abstract: Obstacle avoidance is an important issue in mobile robot localization and navigation. This paper presents a method for mobile robots to autonomously avoid dynamic obstacles and reach the target point, which can be used in Automated Guided Vehicles (AGV), driverless trucks, or other forms of mobile robots for transportation, delivery, and unmanned tasks. This method detects and tracks obstacles by LIDAR, models the obstacle information using a probability grid map. Then, use the advanced Rapidly-exploring Random Tree (RRT) algorithm for local path planning to update the current path to achieve obstacle avoidance based on the obstacle map. The proposed method was implemented in Robot Operating System (ROS) and tested in several real-world environments. This method was able to solve the problem of avoiding dynamic obstacles in real-time during the process of robot autonomous navigation. The average processing time per iteration is 30 ms and 95 ms in the 15m×15m and 50m×50m areas. And the final traveling path is 5% shorter compared with the A* planner and Dynamic Window Approach (DWA) planner used in the Navigation package of ROS.