Local Reference Frames vs. Global Reference Frame for Mobile Robot Localization and Path Planning

This work aims at two major goals: i) to estimate the position of a mobile robot under sensor and control errors; and ii) to provide a motion strategy that outputs a path so that the strategy diminishes the uncertainty about the robot's position both while the robot is moving along the output path and when the robot reaches a goal location. To diminish uncertainty, the path planning strategy switches between reference frames whenever required. We have found out that, for our sensor models, determining the robot position using a set of local reference frames generally yields a smaller position uncertainty than that associated with a global reference frame. To efficiently compute the robot path we use a greedy algorithm in a reduced search space that is able to explore several steps ahead without incurring in too high a computational cost.