Real-time Positioning and Mapping Based on Multi-sensor Fusion for Vehicle System

In the work, we propose a real-time positioning algorithm for vehicle state estimation during localization and mapping process. The method mainly relies on multi-sensor fusion and classical simultaneous localization and mapping (SLAM) technique. A factor map-based optimization framework is firstly designed which tightly couples the measurement produced by LiDAR, IMU and GNSS. New loopback constraints are further added to the back-end optimization problem thereby eliminating cumulative errors. Globally consistent trajectories and mapping results are finally obtained through the proposed method. We demonstrate its better real-time performance and higher localization accuracy compared with existing map-building algorithms by running the algorithm on a Jeston TX2 board.