Path planning between two points for the robot experiencing localization error

We use an approach to simultaneous localization and mapping to determine a path between two points for a mobile robot experiencing localization error in a known environment. Simply instructing a robot to move to a point (x, y)T is not sufficient because the mobile robot’s accrued dead-reckoning error prevents the robot from as-certaining its current location and thus it cannot posi-tion itself at any location, including the goal. If we knew how dead-reckoning error grew, then perhaps we can di-rect the mobile robot to a goal location, module an “error margin. ” Instead, we do not assume an explicit error model, and send the robot to the goal via a sequence of way points, called meet points, whose locations are known a priori. Meet points are nodes of the general-ized Voronoi graph and have the property that a robot can use its range sensors to reliably converge on them via a stable control law. The challenge is then to navi-gate from the final meet point to the goal location (not necessarily on the generalized Voronoi diagram) which constitutes the contribution of this paper. Experiments on the Nomad 200 and soon on the Personal Satellite Assistant Test-bed validate this approach.