Experimental evaluation of a modified obstacle based potential field algorithm for an off-road mobile robot

This article presents an experimental evaluation of a modified obstacle based artificial potential field algorithm for an off-road mobile robot. The first contribution of the presented approach concerns the transformation of the artificial potential field method for the guidance of the vehicle and obstacle avoidance, in order to make it suitable for utilising a visual feedback. The visual feedback is relying on a depth image, provided by the low cost kinect sensor. The second contribution concerns the proposal of a novel scheme for the identification and perception of obstacles. Based on the proposed methodology, the vehicle is capable of categorising the obstacles based on their height in order to alter the calculated forces, for enabling a cognitive decision regarding their avoidance or the driving over them, by utilising the robot's off road capabilities. The proposed scheme is highly suggested for off road robots, since in the normal cases, the existence of small rocks, branches, etc. can be accidentally identified as obstacles that could make the robot to avoid them or block its further movement. The performance of the proposed modified potential field algorithm has been experimentally applied and evaluated in multiple robotic exploration scenarios, where from the obtained results the efficiency and the advantages of such a modified scheme have been depicted.