Collision planner — A probabilistic single stage smooth path planner for mobile robots

Abstract

In this paper a path planning algorithm is proposed for mobile robots. The algorithm utilizes piecewise cubic splines to create a smooth path in the 2D workspace of a mobile robot, while specifying targeted random control points based on the collision detection and penetration characteristics of the candidate paths. The main advantage of the proposed method compared to other probabilistic path planners is that the samples generated are taken only in the free space connecting the start and goal configuration, thus utilizing the geometric characteristics to direct the sampling in proper areas. A wavefront inspired searching algorithm, called collision wave, is used to explore the configuration space, that generates feasible paths that can then be sorted according to a metric. The results show that the directed randomness of the control points reduces the required collision checks and the search space towards high efficiency.