3D Lane Detection with Odometry for High-Speed Vehicle Racing

2026-07-15

AI Summary

This work addresses the failure of conventional lane detection methods in high-speed racing scenarios, where extreme road geometry and highly dynamic environments pose significant challenges. To overcome these limitations, the study introduces odometry pre-integration into 3D lane detection for the first time and proposes a cross-frame geometric modeling approach that fuses multi-camera and inertial measurements. An efficient and lightweight multi-camera integration architecture is also designed. Evaluated on a newly collected high-speed racing multimodal dataset, the method achieves near real-time performance at approximately 300 Hz, with an onboard F1 score exceeding 0.9 and a lateral mean absolute error (MAE) below 0.18 meters in the near-vehicle region. It outperforms existing approaches such as BevLaneDet by improving F1 by 3 percentage points and reducing MAE by over 30%, substantially enhancing robustness and accuracy in high-speed driving conditions.

Abstract

Lane boundary detection is a critical component in autonomous driving systems and has been rigorously studied in regular driving scenarios. However, it is less explored in vehicle racing, where the car moves at higher speeds across more extreme road geometries. To study this problem, we introduce a new dataset for 3D lane detection in racing, featuring >$250$k images from multiple camera feeds and inertial measurements taken with a Lexus LC 500 driving on a closed circuit. With this dataset, we compare various approaches to 3D lane detection and propose modifications that permit frames to be processed at rates of almost 300Hz while retaining high predictive performance in the racing application. This facilitates a multi-camera ensemble approach that is validated on hardware. We show that sensing modalities such as inertial measurements can be leveraged for pre-integration to regress road geometries over both cameras and time, yielding improvements in key metrics. Compared to methods such as BevLaneDet, adding odometry and ensemble predictions improves the F1 score by 3 points and reduces near-vehicle mean absolute errors (MAEs) by $>30 \%$. We show F1 scores $>$0.9 and lateral MAEs of $<$0.18m in vehicle deployments.

Problem

• 3D lane detection
• vehicle racing
• high-speed driving
• road geometry
• autonomous driving

Innovation

• 3D lane detection
• odometry
• multi-camera ensemble
• pre-integration
• high-speed racing