Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Analyzing Infrastructure LiDAR Placement with Realistic LiDAR Simulation Library

Xinyu Cai, Wentao Jiang, Runsheng Xu, Wenquan Zhao, Jiaqi Ma, Si Liu, Yikang Li

TL;DR

The paper addresses optimizing infrastructure LiDAR placement for V2X perception by introducing a Realistic LiDAR Simulation Library (RLS) integrated with CARLA to model 14 LiDAR devices with beam patterns, motion distortion, and ghosting. It proposes a simulate-and-evaluate pipeline and traffic-irrelevant surrogate metrics InfraD and InfraNUC to quickly assess placements, linking point-cloud density and distribution to detection accuracy. Experiments show that, for the same LiDAR count and type, optimized placements can yield about a 15% improvement in average precision in standard lane scenes, and that InfraD and InfraNUC correlate with performance. The work enables faster, more realistic infrastructure LiDAR placement optimization and provides code for reproducibility.

Abstract

Recently, Vehicle-to-Everything(V2X) cooperative perception has attracted increasing attention. Infrastructure sensors play a critical role in this research field; however, how to find the optimal placement of infrastructure sensors is rarely studied. In this paper, we investigate the problem of infrastructure sensor placement and propose a pipeline that can efficiently and effectively find optimal installation positions for infrastructure sensors in a realistic simulated environment. To better simulate and evaluate LiDAR placement, we establish a Realistic LiDAR Simulation library that can simulate the unique characteristics of different popular LiDARs and produce high-fidelity LiDAR point clouds in the CARLA simulator. Through simulating point cloud data in different LiDAR placements, we can evaluate the perception accuracy of these placements using multiple detection models. Then, we analyze the correlation between the point cloud distribution and perception accuracy by calculating the density and uniformity of regions of interest. Experiments show that when using the same number and type of LiDAR, the placement scheme optimized by our proposed method improves the average precision by 15%, compared with the conventional placement scheme in the standard lane scene. We also analyze the correlation between perception performance in the region of interest and LiDAR point cloud distribution and validate that density and uniformity can be indicators of performance. Both the RLS Library and related code will be released at https://github.com/PJLab-ADG/PCSim.

Analyzing Infrastructure LiDAR Placement with Realistic LiDAR Simulation Library

TL;DR

The paper addresses optimizing infrastructure LiDAR placement for V2X perception by introducing a Realistic LiDAR Simulation Library (RLS) integrated with CARLA to model 14 LiDAR devices with beam patterns, motion distortion, and ghosting. It proposes a simulate-and-evaluate pipeline and traffic-irrelevant surrogate metrics InfraD and InfraNUC to quickly assess placements, linking point-cloud density and distribution to detection accuracy. Experiments show that, for the same LiDAR count and type, optimized placements can yield about a 15% improvement in average precision in standard lane scenes, and that InfraD and InfraNUC correlate with performance. The work enables faster, more realistic infrastructure LiDAR placement optimization and provides code for reproducibility.

Abstract

Recently, Vehicle-to-Everything(V2X) cooperative perception has attracted increasing attention. Infrastructure sensors play a critical role in this research field; however, how to find the optimal placement of infrastructure sensors is rarely studied. In this paper, we investigate the problem of infrastructure sensor placement and propose a pipeline that can efficiently and effectively find optimal installation positions for infrastructure sensors in a realistic simulated environment. To better simulate and evaluate LiDAR placement, we establish a Realistic LiDAR Simulation library that can simulate the unique characteristics of different popular LiDARs and produce high-fidelity LiDAR point clouds in the CARLA simulator. Through simulating point cloud data in different LiDAR placements, we can evaluate the perception accuracy of these placements using multiple detection models. Then, we analyze the correlation between the point cloud distribution and perception accuracy by calculating the density and uniformity of regions of interest. Experiments show that when using the same number and type of LiDAR, the placement scheme optimized by our proposed method improves the average precision by 15%, compared with the conventional placement scheme in the standard lane scene. We also analyze the correlation between perception performance in the region of interest and LiDAR point cloud distribution and validate that density and uniformity can be indicators of performance. Both the RLS Library and related code will be released at https://github.com/PJLab-ADG/PCSim.
Paper Structure (10 sections, 2 equations, 6 figures)

This paper contains 10 sections, 2 equations, 6 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Realistic LiDAR Simulation Library
  4. Analyzing Infrastructure LiDAR Placement
  5. Simulate-and-Evaluate with RLS Library
  6. Analysis using Point Cloud Distribution
  7. Experiments
  8. Experimental Setup
  9. Experiments for Infrastructure LiDAR Placement
  10. Conclusions

Figures (6)

  • Figure 1: It is infeasible to evaluate infrastructure LiDAR placements in the real world. Thus, we propose to simulate and then evaluate the placements using the RLS library. We also analyze the correlation between point cloud distribution and perception performance.
  • Figure 2: With the newly proposed RLS library, we can simulate realistic point cloud data using 14 popular LiDAR devices of 3 types, including Surround LiDAR, Solid State LiDAR, and Risley Prism LiDAR. The RLS library can restore the unique characteristics of different kinds of LiDAR.
  • Figure 3: (A) Motion distortion in real point cloud data. (B) Motion distortion simulated by RLS library. Our RLS library simulates the motion distortion effect, which brings realistic point cloud data for moving objects.
  • Figure 4: (A) Ghosting object effect in real point cloud data. (B) A kind of cause of ghosting object effects. (C) A scene simulated by our RLS library.
  • Figure 5: The InfraLOB we set in three simulation scenes. InfraLOB is a virtual region of interest, where we expect infrastructure sensors to perceive the environment precisely.
  • ...and 1 more figures