Repeated Path Planning For Mobile Robots In Uncertain Environments

In most of real-world applications a mobile robot has to drive repeatedly between predefined start and goal points. The path planning has to be done at runtime to avoid unknown obstacles and therefore an actually followed path can be very different from a planned one. This paper describes a method for repeated path planning under uncertainty. The method works by approximating the traversed paths and using them for planning new routes. The approximated path is guaranteed to converge to the goal provided that no obstacles are added or moved during the traversal. A heuristic decision-making strategy is discussed that makes the method suitable for highly dynamic environments. Experiments with a minirobot Khepera suggest that the method quickly adapts to dynamic changes in the environment, reduces travelled distance, travelling time and collision risk