Road Boundary Detection Using 4D mmWave Radar for Autonomous Driving

TL;DR

This work tackles road boundary detection for autonomous driving using a cost-effective 4D mmWave radar. It introduces a three-module pipeline—point cloud preprocessing, point-wise segmentation with a distance-based loss and temporal deviation features, and curve fitting via DBSCAN clustering followed by Gaussian Process Regression—to robustly detect road boundaries in noisy, dynamic environments. The approach demonstrates strong real-world performance, achieving 93% boundary segmentation accuracy and a median Chamfer distance of 0.023 m, with ablation studies confirming the benefits of temporal updating and distance-based penalties. The method shows resilience to varying boundary configurations, noisy scenes, and complex curve shapes, suggesting practical utility for safe navigation and planning in diverse driving scenarios.

Abstract

Detecting road boundaries, the static physical edges of the available driving area, is important for safe navigation and effective path planning in autonomous driving and advanced driver-assistance systems (ADAS). Traditionally, road boundary detection in autonomous driving relies on cameras and LiDAR. However, they are vulnerable to poor lighting conditions, such as nighttime and direct sunlight glare, or prohibitively expensive for low-end vehicles. To this end, this paper introduces 4DRadarRBD, the first road boundary detection method based on 4D mmWave radar which is cost-effective and robust in complex driving scenarios. The main idea is that road boundaries (e.g., fences, bushes, roadblocks), reflect millimeter waves, thus generating point cloud data for the radar. To overcome the challenge that the 4D mmWave radar point clouds contain many noisy points, we initially reduce noisy points via physical constraints for road boundaries and then segment the road boundary points from the noisy points by incorporating a distance-based loss which penalizes for falsely detecting the points far away from the actual road boundaries. In addition, we capture the temporal dynamics of point cloud sequences by utilizing each point's deviation from the vehicle motion-compensated road boundary detection result obtained from the previous frame, along with the spatial distribution of the point cloud for point-wise road boundary segmentation. We evaluated 4DRadarRBD through real-world driving tests and achieved a road boundary point segmentation accuracy of 93$\%$, with a median distance error of up to 0.023 m and an error reduction of 92.6$\%$ compared to the baseline model.

Figures (7)

• Figure 1: 4DRadarRBD System Overview
• Figure 2: (a) Distance loss is calculated as the Euclidean distance between the detected and the actual road boundary points, which has a large value for the noisy points far away from the road boundaries; (b) The deviation vectors for each point in the current frame (e.g., $a_1, a_2, a_3$ for $P_1, P_2, P_3$ respectively) are calculated as the shortest vector from the motion-compensated road boundary points detected in the previous frame to the point in the current frame (e.g., P1, P2, P3).
• Figure 3: Overall Performance of 4DRadarRBD. (a) Confusion matrix for road boundary (RB) point segmentation (accuracy = 93$\%$), (b) Median Chamfer distance error of 4DRadarRBD (our method) and ablation tests without model updating and without distance loss, (c) Median Hausdorff distance error of 4DRadarRBD (our method) and ablation tests without model updating and without distance loss.
• Figure 4: 4DRadarRBD successfully detects road boundaries (referred to as RB) with varying numbers of road boundary curves in: a simple scenario with two curves (top), a forked road intersection with three curves (middle), and a complex urban environment with multiple curves (bottom). In the complex urban area, 4DRadarRBD successfully detects the intersection.
• Figure 5: 4DRadarRBD successfully segments road boundary (RB) points with the noisy points from overpasses and moving vehicles, proving its robustness in complex driving scenarios.