Learning mobile robot's paths using potential field methods

摘要

A mobile robot that navigates in closed environments like houses or offices needs to learn the paths that it can use to cross the rooms. A planner used the approach of artificial potential fields for planning in advance the movements. One of the main problems using this technique is that the planner can stuck into a local minimum where the attraction and repulsion forces cancel each other, thus the movement of the Robot is zero or it oscillates around a certain path. To eliminate this, an expert system uses the knowledge of known obstacles to put intelligently the additional attraction forces in places that will take the robot out of the place where is stuck, specifically in some of the corners of the obstacles. The robot moves between rooms to reach its destination and larns a new path found by the potential algorithm.