Towards Resolving the Kidnapped Robot Problem: Topological Localization from Crowdsourcing and Georeferenced Images

Abstract

In this research, we address the kidnapped robot problem, a fundamental localization problem where a robot has been carried to an arbitrary unknown location for which no prior maps exist. In our approach, topological maps are created using a single sensor, namely, a camera, with the aim to localize the robot and drive it to its initial, home, position. Our approach differs significantly from other approaches that attempt to solve this localization problem. In order to localize a robot within an unknown environment we exploit the potential of social networks and extract the GPS information from in georeferenced images. The experiments carried out within a university campus affirm the validity of our approach and provide the means to resolving similar problems with the methods presented.