Improved Hector-SLAM Algorithm Based on Data Fusion of LiDAR and IMU for a Wheeled Robot Working in Machining Workshop

Simultaneous localization and mapping (SLAM) is the key technology to achieve efficient operation of mobile robots, especially in complex production workshops, logistics warehouses and other typical intelligent manufacturing scenarios, robots must first rely on accurate prior map information to achieve autonomous movement, so that intelligent processing, assembly, human-machine collaborative operation and other tasks can be performed. However, maps are often built depending on expensive LiDAR, which greatly increases costs and makes it difficult to popularize. Therefore, this paper deploys a wheeled robot turtlebot2 equipped with low-cost single-line LiDAR hokuyo (UTM-30LX) to build two-dimensional grid maps in the machining workshop. According to the characteristics of the robot and the scene, three laser SLAM algorithms are configured based on ROS (kinetic) under ubuntu 16.04 LTS system, including GMapping-SLAM, Hector-SLAM and Cartographer-SLAM. Compared with other two algorithms, Hector-SLAM algorithm has better performance in terms of the completeness of the map restoration scene. However, there are drift and overlap problems in the local map. To solve these problems, a map building method, improved Hector-SLAM algorithm, is proposed which first eliminates the invalid laser data and then uses inertial measurement unit (IMU) to compensate. Comparative experiments show that the method effectively improves the accuracy of the map and improves the work efficiency of the robot.