VLPV: Virtual LiDAR-based Point Cloud Processing Verification via Simulation Environments

Effective verification of vehicle-mounted LiDAR point cloud processing algorithms is a critical step in ensuring the safety and reliability of autonomous driving systems. Traditional real-world data-based verification methods suffer from several inherent limitations, such as high costs, difficulties in scenario reproduction, and the lack of precise ground truth. To address these challenges, this paper proposes a virtual simulation-based verification method for vehicle-mounted LiDAR point cloud processing. Leveraging high-fidelity virtual environments and physical sensor models, this method employs techniques such as ray tracing to generate synthetic LiDAR point clouds and associated sensor data with precise and synchronized ground truth. This paper details the verification workflow, key technical components, and its application in point cloud processing validation. Experiments demonstrate that the method establishes a looped verification process incorporating similarity comparison, preprocessing, and optimization, systematically evaluating the consistency between virtual and real-world data. This provides a low-cost, high-efficiency, reproducible, and ground-truth-supported approach for the development of autonomous perception algorithms, thereby significantly advancing LiDAR point cloud algorithm development and performance evaluation.