Triangle-based loop closure detection method for reflector mapping and localization for logistics robot

Purpose The purpose of this paper is to propose a loop detection method tailored to mapping and localizing reflectors for logistic robots. A crucial component, a shape feature construction method, is essential to fulfill loop requirements. Furthermore, the demand for map optimization procedures aligned with the system is also emphasized. Design/methodology/approach A differential chassis designed for navigation within a workshop serves as an information collector, gathering data from LiDAR, odometer and reflectors. The authors introduce a novel triangle construction method for mapping and localizing logistics robots. Additionally, a sub-map optimization procedure based on triangles has been developed to enhance mapping and localization accuracy. Findings A triangle shape-based loop detection method greatly improves the positioning accuracy of reflectors across sub-maps. The graph-based optimization procedure proves to be effective, especially in scenarios featuring a long corridor and annular space. The discovery of an effective shape matching method has been validated as beneficial for mapping and locating the robot. Research limitations/implications If only limited information is available, then the effectiveness of the backend optimization will be impacted. Randomly distributed spatial features are used to replace the artificial reflectors to enhance the generalization ability of the entire system. This study introduces dynamic features and promotes the capability of mapping the dynamic scene. Practical implications A triangle feature construction method for logistics robots is designed; besides, the internal and inter-constraints between triangles are constructed. A novel loop detection method has been proposed, which determines the loop-back position based on triangle similarity. Additionally, a novel octree-based sub-map store and retrieve method has been introduced; by use this method, the loop detection time can be complete within 10 ms. Social implications The triangle-based method can be extended into other domain, by introducing other distinct features as loop detection descriptors. Besides, additional attribute can also be incorporated into the loop descriptor to further enrich and refine the method. Originality/value This paper introduces a triangle construction method that forms the basis for a loop detection technique aimed at establishing relationships between sub-maps. In addition, a graph-based optimization procedure is used to generate an accurate map of reflectors. The shape-based loop detection method proposed in the paper is beneficial for the current system and may offer an intuitive solution for similar mobile robots.