Robot navigation in hand-drawn sketched maps

Many navigation systems for mobile robots use a metrical map for localization and trajectory planning. In several situations, however, building such a map upfront is hard or even impossible. In this paper, we present a combined approach for robot localization and navigation solely relying on a hand-drawn sketch of the environment. In our work we model the sketch as a two-dimensional manifold with an unknown metric. We use Monte Carlo Localization for estimating the state of the robot in combination with the local deformation of the sketch. Accordingly, our approach can track both, the pose of the robot and the local metric properties of the drawing. Furthermore, we couple the pose tracker with a path planner that uses the global information of the sketch and the estimated local metric to robustly perform navigation tasks in real world environments.