A Navigation Framework Fused with 3D and 2D SLAM Algorithms for solid-state LiDARs

Abstract

Nowadays, most navigation frameworks for mobile robots are designed based on mechanical LiDARs, and there are relatively few navigation frameworks for solid-state LiDARs with small FoV(Field of view) which have been widely used in recent years because of its lower cost and long-range. In this paper, we present a navigation framework for solid-state LiDARs with small FoV like Livox. This framework is based on the ROS system and combines 3D SLAM algorithm Loam_Livox with 2D SLAM algorithm Gmapping. The odometer information output by Loam_Livox is used as input for the occupancy grid map construction of algorithm Gmapping. The occupancy grid map information output by algorithm Gmapping is also passed into the move_base framework to control the mobile robot for the navigation function. In the move_base framework, we use improved A-STAR and improved DWA algorithms, which can improve the working safety and efficiency of the mobile robot. As a result, this system is able to build 3D dense point cloud map and occupancy grid map in real time, which can also perform navigation tasks for the mobile robot.