Heuristic search approach for mobile robot trap recovery

Abstract

Abstract Trap recovery is required for autonomous mobile robots that sense and avoid local obstacles as they maneuver along a goal-oriented path. We have developed a heuristic search-based recovery algorithm that isinvoked immediately whenever a potential trap situation is identified. This algorithm generates intermediate short distance target points (via points) that are used as temporary goals to guide the robot out of thetrap situation and allow it to reach a final destination without stopping-then-thinking whenever traps are encountered. Real-time navigation is achieved by reacting quickly to the local sensor-detected feedbackinformation and employing on-line reasoning and planning to provide trap recovery. I. Introduction It is often important for mobile robots to navigate autonomously in unknown and changing environments.Several local obstacle avoidance algorithms have been developed for robot guidance [1, 2, 4, 5]. However, mostof them do not provide recovery from simple a-shaped trap situations (e.g. dead end) which is vital for trulyautonomous navigation. We have developed an advanced mobile robot navigation system that can be integratedinto the control systems of mobile robots to provide improved autonomous guidance. The techniques describedhere have been used with a modified commercial mobile robot platform that is intended to provide assistance toindividuals with mobility, cognitive and perceptual difficulties as well as to perform various tasks in hazardousenvironments.We present an approach which allows the mobile robot to recover from trap situations. This new approachdevelops two maps concurrently, a high level map for local obstacle avoidance and a low level map for traprecovery. A local obstacle avoidance algorithm used with the high level map allows the robot to traverseits pathand avoid collision with enroute obstacles. Whenever the robot nears or enters a trap, a heuristic search-basedrecovery algorithm is invoked to compute intermediate via points1 as temporary targets that move the robot awayfrom the trap. The trap recovery algorithm uses the low level inap and employs heuristic rules to determine thesearch goal and to evaluate the qualifications of each candidate via point. This approach allows the algorithmto employ environmental knowledge to accomplish more intelligent navigation. The robot continues its journeyunder the guidance of the local obstacle avoidance algorithm by using either the nearest qualified via point or thefinal target as its heading direction.The experimental results clearly demonstrate the success of this new approach. Because of the utilization of thetechniques described in the next section, our mobile robot recovers from trap situations without stopping-then-