A brief review of simultaneous localization and mapping

Abstract

This paper reviews the development history of simultaneous localization and mapping (SLAM) and concentrates on two mainstream methods: the filter-based method and the vision-based graph optimization method. FastSLAM and Real-Time Appearance-Based Mapping (RTAB-MAP) as two examples are adopted in the real experiments. The experiments are implemented on TurtleBot with Kinect in a small laboratory and a large circular corridor. The experimental results show that the error is small in the small laboratory, but in the highly unknown large scale environment, the loop closure detection is less effective and the error accumulation is obvious. The results show the accuracy and robustness of two algorithms need to be further improved when robots are in the large-scale unknown environments.