Rapidly-exploring Random Trees Based on Cylindrical Sampling Space

Abstract

In this paper, the defect of traditional fast-expanding random tree (RRT) is optimized for the path planning in three-dimensional environment. Due to the randomness generated by its own nodes, the traditional RRT algorithm has the defects of the planned path twist and easy to cross the target point, and greatly differs from the ideal path, so that the time required for the robot to implement path tracking is greatly increased. In this paper, an RRT algorithm based on cylindrical sampling space is proposed. The algorithm uses the line connecting the starting point and the ending point of the path as the central axis. The maximum distance from all nodes on the path to the cylinder axis is taken as the radius value of the first cylinder. The cylinder space is used as the new node sampling space in the next path planning. Then, a new cylinder is generated by reducing the radius of the previous cylinder as the current sample sampling space. Repeat the above steps until find the last optimized path. The path generated by the traditional RRT algorithm and the path generated by the RRT optimization algorithm based on cylindrical sampling space are compared and analyzed under the given obstacle and obstacle-free environment. The simulation results show that the path generated by the RRT optimization algorithm based on the cylindrical sampling space is smoother and shorter, almost no redundant path. The path meets the actual motion requirements of the robot.