Path planning in extended uncertain environments

This paper presents a new approach to the problem of planning the motion of a mobile robot in extended uncertain environments. All knowledge of the environment has been acquired by the robot during the current (or a previous) operation, such that the environment description reflects the accumulated error of the robot's pose during periods of dead-reckoning navigation. In this uncertain environment, the robot searches for trajectories that maximize the probability of attaining a desired target region. For that purpose we identify a discrete set of robot positions in order to construct a routing graph, whose arcs represent the probability of reaching a new position. In that way the search for an optimal trajectory is solved by searching for a minimum weight path in a routing graph. The method is based on a probabilistic model of all the errors/uncertainties affecting the reliability of the planned trajectory.