Investigation of Camera-Based Loop-Closing Techniques for EKF-SLAM

This paper deals with the implementation and an assessment of loop-closing techniques for an EKF-SLAM. Here, especially the minimization of uncertainties of the vehicle pose and of the global map are in the scope of interest. Therefore, different approaches of loop-closing are integrated in the framework of an extended Kalman Filter implementation for the SLAM problem. In particular, the detection of a loop closure as well as its impact on the accuracies of both robot pose and global map will be presented. Experimental results are obtained from a Turtlebot3 robot operating in a lab environment.