Zhang, Chaokun (Tianjin University), Zhou, Bolin (Tianjin University), Lv, Aojia (Tianjin University)

LiDAR Localization System of Intelligent Connected Vehicle in Low-Texture Parking Garages

Scheduled for presentation during the Poster Session "Accurate Positioning and Localization" (ThAT17), Thursday, September 26, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Real-time and accurate localization is a pivotal aspect of the Intelligent Connected Vehicle (ICV) system. Traditional LiDAR localization techniques encounter significant challenges in low-texture and long-tailed indoor parking garages, primarily due to the scarcity of matching feature points, which hinders the attainment of sufficient constraints and stable optimization. To address this problem, we have devised a multifaceted approach. Firstly, We introduce a novel intensity-based Referable Static Object (RefSO) segmentation strategy. This strategy effectively filters out the potential dynamic objects and ground points from the point cloud, thereby facilitating the extraction of skeleton points from stable features. Secondly, combined with the long-tailed environment of the parking garage, we propose a feature extraction method based on the Field of View (FoV) segmentation, and the intensity is added to the odometry to enhance the matching accuracy. Lastly, we have implemented an indoor parking garage-specific vehicle localization system based on the priori map. The efficacy of our system is validated through real-world data we collected. The localization result of ICV using Robosense 16 (RS-16) is significantly improved, with a global drift of just 0.013% even with sparse point clouds.