Which LiDAR scanning pattern is better for roadside perception: Repetitive or Non-repetitive?

TL;DR

This work addresses how LiDAR scanning patterns affect roadside perception by introducing InfraLiDARs' Benchmark, a CARLA-based dataset and framework for comparing repetitive versus non-repetitive scanning. It deploys a two-pronged evaluation: a statistical benchmark of scan characteristics and a perception benchmark across four state-of-the-art 3D detectors (PointRCNN, PointPillars, PV-RCNN, DSVT) in highway, crossroad, and curve scenarios. The key finding is that non-repetitive and 128-line repetitive LiDARs deliver comparable detection performance, with non-repetitive offering superior long-range detection at a lower cost but a smaller horizontal field of view. The study provides data-driven deployment recommendations and publicly releases the InfraLiDARs' Benchmark to catalyze future roadside-perception research, while acknowledging simulated-environment limitations and the need for broader scenario validation.

Abstract

LiDAR-based roadside perception is a cornerstone of advanced Intelligent Transportation Systems (ITS). While considerable research has addressed optimal LiDAR placement for infrastructure, the profound impact of differing LiDAR scanning patterns on perceptual performance remains comparatively under-investigated. The inherent nature of various scanning modes - such as traditional repetitive (mechanical/solid-state) versus emerging non-repetitive (e.g. prism-based) systems - leads to distinct point cloud distributions at varying distances, critically dictating the efficacy of object detection and overall environmental understanding. To systematically investigate these differences in infrastructure-based contexts, we introduce the "InfraLiDARs' Benchmark," a novel dataset meticulously collected in the CARLA simulation environment using concurrently operating infrastructure-based LiDARs exhibiting both scanning paradigms. Leveraging this benchmark, we conduct a comprehensive statistical analysis of the respective LiDAR scanning abilities and evaluate the impact of these distinct patterns on the performance of various leading 3D object detection algorithms. Our findings reveal that non-repetitive scanning LiDAR and the 128-line repetitive LiDAR were found to exhibit comparable detection performance across various scenarios. Despite non-repetitive LiDAR's limited perception range, it's a cost-effective option considering its low price. Ultimately, this study provides insights for setting up roadside perception system with optimal LiDAR scanning patterns and compatible algorithms for diverse roadside applications, and publicly releases the "InfraLiDARs' Benchmark" dataset to foster further research.

Figures (13)

• Figure 1: Visualization of different LiDAR point clouds. The three images in the upper section depict point cloud patterns from repetitive scanning LiDARs, while those in the lower section depict patterns from non-repetitive scanning LiDARs.
• Figure 2: Framework of Benchmarking
• Figure 3: The graph illustrates the velocity distribution of the three scenes. The red numerals above the chart indicate the count of vehicles whose speed exceeded the maximum value displayed on the x-axis.
• Figure 4: Lidar position and orientation
• Figure 5: Visualization of point clouds of different LiDARs