Path Planning by Intelligent Autonomous Robotic Vehicles with Growing World Models

Abstract The problem of intelligent path planning by autonomous robotic vehicles in unordered environments is considered for the case where the two restrictions are imposed at a time: a) the world is unknown and must be modelled by the robot on the basis of sensory data; b) only local sensory information in a very limited vicinity of a current location of the robot is available. An approach to non-heuristic obstacle avoidance and path generation based upon self-learning is presented. Algorithms for active formation of world models are described that permit the robot or a team of interacting robots to fuse “local knowledge” in a growing graph model in order to plan rational (eventually optimal) paths.