Loop Closures in LiDAR-based SLAM systems

Abstract

SLAM technology stands for Simultaneous Localization and Mapping. It is a process that uses sensors mounted on a robot to construct a map of an unknown environment while simultaneously keeping track of its own location within the map. Performing both localization and mapping simultaneously is a complex problem because the two pro- cesses are completely dependent of each other. Sensors provide information, but they always have some inaccuracy and noise which makes the SLAM process more complex. Therefore, a SLAM system su ers from long-term distance measurement drift, and the only way to overcome this issue consists of using loop closure detection algorithms to correct the drift error and build a globally consistent map of the environment. To do so, a loop closing technique based on Scan-Context descriptor is studied and compared to a conventional approach based on Scan-Matching, and it shows a su ciently improved performance. Also, we investigate some recent SLAM implementations (both in ROS and in Matlab 2021b) commonly evaluated in indoor/outdoor environments.