A simple experimental method to estimate and benchmark automotive LIDARs performance in fog

Abstract

LiDARs hold promise for various automotive applications, but their performance in adverse weather conditions remains a severe limitation. Indeed, fog can compromise the ability to perform fundamental tasks such as detection, classification, and tracking. The success of these tasks depends on the quality of the data provided by the LiDAR, i.e., the point cloud, PC, and the algorithms used to analyse that PC. Some previous studies exploited large and sophisticated facilities filled with fog to analyse LiDARs in fog. However, such facilities are intrinsically highly complex and costly. To overcome these limitations, we propose a much less expensive method based on a fog chamber, a 1 m side transparent chamber to be placed between the LiDAR and the targets, then filled with fog. The proposed method allows for the analysis and comparison of the performance of both LiDARs and processing algorithms while avoiding the cost and complexity of large facilities and the limitations intrinsic to mathematical modelling and numerical simulation. To provide examples of the information that is obtainable using the proposed method, the results from a popular LiDAR and processing algorithm, namely, the Velodyne VLP 16, and the MATLAB^® Computer Vision Toolbox, are also reported.