Physical-consistent behavior embodied in B-spline curves for robot path planning

In this paper, a path planning algorithm for smooth obstacle avoidance of mobile robots is presented and discussed. The idea is based on B-splines, which are trajectories defined by a linear combination of basis functions weighted by constants called control points. Since the overall curve depends locally only on the position of a limited set of control points, B-splines are ideal tools for generating online reference trajectories able to react to local changes of the environment (e.g. mobile or spawning obstacles). The via-points of the reference B-spline are treated as dynamic agents with customized dynamic properties. These agents interact with each other and with the environment in such a way to generate the correct reference for the B-spline generator and thus for the robot. The high flexibility and the low computational burden of the proposed algorithm allow to develop applications in partially unknown or dynamically changing environments since a local modification of the trajectory does not require an entire re-computation. Experimental results performed with a KUKA Youboot mobile robot within a ROS based set-up are reported.