2D-Span Resampling of Bi-RRT in Dynamic Path Planning

Abstract

Path planning is an essential task in robot soccer to enable the robot to quickly arrive at a desired location from which it can shoot or dribble the ball to a goal. Previous work in path planning used sonar or laser-based sensors to obtain local information for avoiding obstacles and reaching the goal. In the process, the robot may move slowly and collide easily with other robots using similar obstacle-avoidance algorithms. This work proposes a 2D-span resampling method and post processing including pruning and smoothening of bi-directional rapidly-exploring random trees (Bi-RRT) to improve the path route and computational time of path planning. To avoid obstacles, the path is re-planned using a novel 2D-span resampling method in Bi-RRT. The post processing of pruning unnecessary Bi-RRT nodes and smoothing the path route enables a robot to reach the goal via a shorter path. Simulations showed the proposed method outperformed several common path-planning methods, generally resulting in a shorter route distance and less computational time.