Towards exteroceptive based localisation

The intelligent application of a mobile robot, outside the experimental laboratory, requires a robust locomotive strategy that is rarely conducive to stringent kinematic modeling. Localisation methods that rely upon such modeling often fail, as model boundaries succumb to unpredictable events. This paper presents the development of a self-contained localisation system that purposely obviates the need for odometric information, and an associated kinematic model, to provide robot anonymity. Without odometry, the system is oblivious to the non-systematic vagaries of the robotic platform interacting with a natural domain. The proposed system hypothesises about the robot's absolute pose by algorithmically solving the kidnapped robot problem using exteroceptive based perception. Since no a priori information is assumed, long-term pose fixes are derived within a simultaneous localisation and mapping (SLAM) framework. Preliminary results were gathered using a skid steering mobile robot, equipped with a scanning laser rangefinder, in an outdoor environment. This novel localisation approach was found to be efficient and robust, while exhibiting the capacity for widespread applicability.