Lightweight SLAM with automatic orientation correction using 2D LiDAR scans

Abstract

Simultaneous localization and mapping (SLAM) is about consistent maps in the long run. Loop closing is the most popular way for ensure long-term consistency in presence of multiple measurements by the same or multiple robots. Loop closure can be executed using raw odometrical data, but a more sophisticated, yet still light-weight method is presented in this paper: a landmark descriptor-based relative displacement calculation method for diminishing unwanted orientation errors that otherwise often lead to map inconsistency. Landmark descriptors are created using light detection and ranging (LiDAR) scans and the relation is calculated using scan-matching. The novelty of this research is a method providing long-term orientation and position correction without additional overhead between landmark detections, thus enabling simple agents to do the SLAM in a cooperative way.