A Novel Perception Entropy Metric for Optimizing Vehicle Perception with LiDAR Deployment

TL;DR

This work tackles the challenge of fast, accurate evaluation and optimization of LiDAR-based vehicle perception. It introduces a novel perception-entropy metric, PE-VGOP, based on vehicle grid occupancy to rapidly reflect how point-cloud distributions influence vehicle detection, and it couples this metric with a Gazebo-based LiDAR deployment simulator. A DE-PSO optimization framework is developed to maximize the aggregate perception entropy by tuning LiDAR placement and tilt, with validation on KITTI showing strong alignment with ground-truth detection (correlation > $0.98$) and field tests demonstrating up to $25\%$ Recall improvements for an RS-32 LiDAR. The approach provides a practical, distribution-aware tool for designing and evaluating LiDAR deployments across different beam configurations, with potential for broader beam-distribution analyses and future sensor-design insights.

Abstract

Developing an effective evaluation metric is crucial for accurately and swiftly measuring LiDAR perception performance. One major issue is the lack of metrics that can simultaneously generate fast and accurate evaluations based on either object detection or point cloud data. In this study, we propose a novel LiDAR perception entropy metric based on the probability of vehicle grid occupancy. This metric reflects the influence of point cloud distribution on vehicle detection performance. Based on this, we also introduce a LiDAR deployment optimization model, which is solved using a differential evolution-based particle swarm optimization algorithm. A comparative experiment demonstrated that the proposed PE-VGOP offers a correlation of more than 0.98 with vehicle detection ground truth in evaluating LiDAR perception performance. Furthermore, compared to the base deployment, field experiments indicate that the proposed optimization model can significantly enhance the perception capabilities of various types of LiDARs, including RS-16, RS-32, and RS-80. Notably, it achieves a 25% increase in detection Recall for the RS-32 LiDAR.

Figures (9)

• Figure 1: LiDAR deployment optimization framework.
• Figure 2: Simulator for LiDAR deployment and vehicle point cloud collection.
• Figure 3: Visualization of vehicle coordinate system.
• Figure 4: Example of VGOP with the grid dimensions $\mu_t=\mu_s=\mu_f=0.05 \text{m} \times 0.05 \text{m}$.
• Figure 5: Comparison of the proposed PE-VGOP and SOTA metrics in evaluating LiDAR perception performance.