CuikSLAM: A Kinematics-based Approach to SLAM

摘要

In this paper, we depart from the fact that Simultaneous Localization and Mapping (SLAM) is a sub-case of the general kinematic problem, and, thus, all techniques used in kinematics are potentially applicable to SLAM. We describe how to formalize a SLAM problem as a typical kinematic problem and we propose a simple SLAM algorithm based on an interval-based kinematic method called Cuik previously developed in our group. This new algorithm solves the SLAM problem taking advantage of the structure imposed in the SLAM problem by the motion and sensing capabilities of the autonomous robots. However, since we use a kinematic approach instead of a probabilistic one (the usual approach for SLAM) we can perfectly model the constraints between robot poses and between robot poses and landmarks, including the nonlinearities, and we can ensure those constraints to be fulfilled at any time during the map construction and refinement. The viability of the new algorithm is shown with a small test.