Collision Avoidance for Dynamic Obstacles Control via Navigation Function

Abstract

This paper deals with an obstacle avoidance problem based on a potential field method for mechanical systems. The potential field method using a navigation function is effective for obstacles avoidance control, since this function has no local minima and the derived control input is bounded on obstacles' edges. However, if there exist dynamic obstacles which move quickly, the robot using this method will collide with them. We propose a new potential function based on a navigation function, which varies according to the velocities of the dynamic obstacles. Finally, we show the effectiveness of our method by a simulation.