Proposed neuro-guided learning for obstacle avoidance in AMBO a robotic device

Abstract

Mobile robots, utilized increasingly in many applications, require accurate methods of obstacle avoidance. The proposed method assumes the robotic devices operational mechanism does not require data in the form of obstacle recognition, of the obstructions encountered. The exploration of relatively unknown environments including aerial, undersea, desert, icescapes or any dynamic environments in most instances require that obstacle avoidance be both automatic and autonomous and the matters of obstacle recognition (OR) may then be left to the controller, observer or to higher level systems where algorithms for visual or other recognition mechanisms may be achieved. A Guided Learning algorithm has been selected for evaluation to be incorporated within the robot system to allow high speed, memory-like reactions to its manoeuvres within chaotic, obstruction laden environments. The robotic device consists of a quad track quad motor crawler type vehicle, purpose built to serve the function of an autonomous environmental research vehicle and as such the paper deals with the portion of the system relating to purely OA matters. The device utilizes the received sensor data from a 24 segment passive array combined with a Guided-Learning system to control the general motion of the machine through remote and unknown locations.