Simulating Realistic LiDAR Data Under Adverse Weather for Autonomous Vehicles: A Physics-Informed Learning Approach

arXiv cs.RO / 4/3/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper argues that existing LiDAR simulation methods fall short under adverse weather because they cannot accurately model how atmospheric phenomena affect LiDAR signal propagation and geometry.
It introduces a physics-informed learning framework (PICWGAN) that combines physics-driven constraints (e.g., signal attenuation and geometry-consistent degradations) with generative modeling to synthesize more realistic weather-affected LiDAR data.
Experiments on real-world datasets for snow (CADC) and rain (Boreas), along with the VoxelScape dataset, show that the generated LiDAR intensity patterns better match real-world measurements.
The authors evaluate realism using multiple quantitative metrics (MSE, SSIM, KL divergence, and Wasserstein distance) and report statistically consistent intensity distributions compared with real data.
They find that perception models trained on PICWGAN-enhanced synthetic data outperform baselines in downstream 3D object detection, reaching performance comparable to models trained on real-world LiDAR under adverse weather.

Abstract

Accurate LiDAR simulation is crucial for autonomous driving, especially under adverse weather conditions. Existing methods struggle to capture the complex interactions between LiDAR signals and atmospheric phenomena, leading to unrealistic representations. This paper presents a physics-informed learning framework (PICWGAN) for generating realistic LiDAR data under adverse weather conditions. By integrating physicsdriven constraints for modeling signal attenuation and geometryconsistent degradations into a physics-informed learning pipeline, the proposed method reduces the sim-to-real gap. Evaluations on real-world datasets (CADC for snow, Boreas for rain) and the VoxelScape dataset show that our approach closely mimics realworld intensity patterns. Quantitative metrics, including MSE, SSIM, KL divergence, and Wasserstein distance, demonstrate statistically consistent intensity distributions. Additionally, models trained on data enhanced by our framework outperform baselines in downstream 3D object detection, achieving performance comparable to models trained on real-world data. These results highlight the effectiveness of the proposed approach in improving the realism of LiDAR data and enabling robust perception under adverse weather conditions.