Clustering Based Loop Closure Technique for 2D Robot Mapping Based on EKF-SLAM

Abstract

Simultaneous Localization and Mapping(SLAM) is an important technique to realize autonomous navigation of a mobile robot in an unknown environment. The SLAM problem involves a mobile robot to continuously take measurements using sensors, localize its position in the environment and simultaneously built a map of the environment it has visited. For any previously visited environment the system must be able to calculate the relative transformation between the measured and predicted states also called as Loop Closure. In this paper, we propose clustering based techniques for realizing fast loop closure for indoor robot mapping. While utilizing the standard Extended Kalman Filter(EKF) based SLAM algorithm, we propose clustering techniques for finding landmarks for realizing Loop Closure. Through experimental results the proposed algorithm is found to be simple and robust enough for faster loop convergence for SLAM problem.