Accurate Cooperative Localization Utilizing LiDAR-equipped Roadside Infrastructure for Autonomous Driving

TL;DR

This work tackles robust autonomous vehicle localization in environments lacking distinctive map features by introducing a cooperative localization framework that leverages LiDAR-equipped roadside infrastructure (RSUs) and V2I communications. Built on Autoware, the system enables RSUs to estimate the vehicle's 2D pose from roadside LiDAR and CAM-provided vehicle dimensions, then fuse this information with onboard EKF/NDT localization to achieve higher accuracy. Key contributions include an improved roadside LiDAR perception pipeline, an EKF-based data fusion strategy, and end-to-end evaluation in the AWSIM Nishishinjuku environment showing substantial localization improvements (60–80%) and resilience to network delays and packet loss. The approach enhances reliability and scalability of autonomous driving in challenging urban scenarios by extending sensing infrastructure and reducing dependence on onboard-only localization.

Abstract

Recent advancements in LiDAR technology have significantly lowered costs and improved both its precision and resolution, thereby solidifying its role as a critical component in autonomous vehicle localization. Using sophisticated 3D registration algorithms, LiDAR now facilitates vehicle localization with centimeter-level accuracy. However, these high-precision techniques often face reliability challenges in environments devoid of identifiable map features. To address this limitation, we propose a novel approach that utilizes road side units (RSU) with vehicle-to-infrastructure (V2I) communications to assist vehicle self-localization. By using RSUs as stationary reference points and processing real-time LiDAR data, our method enhances localization accuracy through a cooperative localization framework. By placing RSUs in critical areas, our proposed method can improve the reliability and precision of vehicle localization when the traditional vehicle self-localization technique falls short. Evaluation results in an end-to-end autonomous driving simulator AWSIM show that the proposed method can improve localization accuracy by up to 80% under vulnerable environments compared to traditional localization methods. Additionally, our method also demonstrates robust resistance to network delays and packet loss in heterogeneous network environments.

Figures (7)

• Figure 1: The structure of the proposed system and experimental setup. Our proposed method involves the implementation of a roadside localization module, which includes conventional vehicle position estimation using roadside LiDAR and a position refinement algorithm, detailed in Section \ref{['sec:improvement']}. To evaluate the performance of the proposed method, we conducted experiments in a virtual Nishishinjuku environment using the AWSIM simulator.
• Figure 2: Position refinement algorithm utilizing ground truth vehicle dimensions. Vehicle length and width data are transmitted from the CAV to the RSU via V2I communication. By aligning ground truth vehicle box to the alignment point, the RSU can estimate a more accurate vehicle pose.
• Figure 3: This figure illustrates the point filtering process used in our position estimation algorithm. Points are sorted by height, and only the lower ones are retained for further analysis. In the visualization, colored points represent those used in the algorithm, while white points—including both background points and higher vehicle points—are filtered out.
• Figure 4: Illustration of the localization error evaluation conducted on the Nishishinkjuku map Matsumoto2023-wr. Locations A and B were selected as the experimental sites for our study.
• Figure 5: Comparison of localization errors between the baseline and proposed methods using two different LiDAR models.