Tightly-Coupled Lidar-Based Lifelong Mapping using Wheels and a MEMS Gyroscope for Mobile Robots in Dynamic Environment

摘要

Localization and mapping technology of mobile robots is important in production and warehousing. The changing warehousing environment always result in poor accuracy and robustness of the mobile robot's localization. It is necessary to detect the degree of environmental changes and update the map. This paper proposes a novel lifelong mapping method fusing measurements from the lidar, wheels and a MEMS gyroscope, which can create a reliable map corresponding to the current working environment. Firstly, we use the wheels and gyroscopes measures for pre-integration to provide relative motion constraints, while integrating wheels and gyroscopes to provide initial pose for lidar matching. We also establish lidar odometry by submaps to provide another type of relative motion constraints. The absolute constraints are obtained by matching the lidar measurements with prior map. These three types of constraints are optimized in a factor graph. At the same time, during the process of updating the map, detect newly added artificial markers such as reflective columns or strips, and update them into the map timely. Then, in order to delete redundant lidar data keyframes for building map, we propose a determine redundancy method based on submaps. Finally, we propose a method for calculating the degree of map changes, which can update the map timely. Experimental results prove that our method can establish higher quality maps, optimize lidar external parameters timely during the mapping process, delete redundant lidar keyframes more efficiently, and update the current environmental map more accurately and timely.