Loop-Closure Detection With a Multiresolution Point Cloud Histogram Mode in Lidar Odometry and Mapping for Intelligent Vehicles

Abstract

Precise positioning is the basic condition for intelligent vehicles to complete perception, decision making and control tasks. In response to this challenge, in this article, lidar simultaneous localization and mapping (SLAM) is taken as the research object, and a SLAM system is designed that integrates motion compensation and ground information removal functions, and can construct a real-time environment map and determine its own position on the map while the vehicle is driving. A loop-closure detection method with a multiresolution point cloud histogram mode is proposed, which can effectively detect whether the vehicle passes through the same position and perform optimization to obtain globally consistent pose and map information in the urban conditions with more driving loops. We conduct experiments on the well-known KITTI dataset and compare the results with those of state-of-the-art systems. The experiments confirm that the lidar SLAM system designed in this article can provide accurate and effective positioning information for intelligent vehicles. The proposed loop-closure detection algorithm has an excellent real-time performance and accuracy, which can guarantee the long-term driving operation of these vehicles.