LiDAR-Based NDT Matching Performance Evaluation for Positioning in Adverse Weather Conditions
Jiachong Chang, Runzhi Hu, Feng Huang, Dingjie Xu, Li‐Ta Hsu
Abstract
Light detection and ranging (LiDAR) can provide continuous and stable pose estimation with the model of normal distribution transform (NDT), which is widely used in autonomous vehicles (AVs), even under adverse weather conditions. However, there are few studies about the influence of inclement weather on LiDAR positioning results. In this article, different weather scenarios (rain, fog, and snow) are composed of synthetic LiDAR datasets based on state-of-the-art weather simulators. Then, the impacts of different adverse weather conditions are quantitatively evaluated in terms of positioning accuracy and uncertainty. Afterward, we perform the first study to qualitatively analyze the relationship between meteorological weather standards and LiDAR positioning performances, which is significant but unexplored. Evaluated results indicate that NDT matching performance will deteriorate in adverse weather conditions, especially when the meteorological level is “Heavy” or “Violent,” threatening the AVs’ positioning security seriously. Therefore, the results of this article provide more basis for the realization of high-precision positioning in adverse weather conditions, to ensure the positioning safety of AVs.