Performance Evaluation Method of LiDAR in Rainfall Conditions

Automotive LiDAR has become an important environment sensing sensor used in the autonomous driving system (ADS) due to its advantages of large detection range, high accuracy, and high resolution. However, due to the constraints of harsh weather conditions such as rainfall, the inability of LiDAR to maintain excellent detection performance in all operating conditions has become a major limiting factor in its technological application. Therefore, it is necessary to test and verify the function and safety of the LiDAR in the rainy environment to support the adjustment of the perception strategy of the ADS. Currently, vehicle OEMs typically test the performance of LiDAR in actual dynamic scenarios in natural rainfall environments. The testing process is not reproducible and cannot quantitatively analyze the impact of rainfall on LiDAR. On the other hand, the direction of performance improvement by LiDAR manufacturers is weakly related to the testing scenario. In order to comprehensively evaluate the issues in the test scenario and determine improvement directions of LiDAR, the performance evaluation method of LiDAR based on the safety of the intended functionality (SOTIF) scenario deconstruction in rainfall conditions is proposed. The risk factors of LiDAR were extracted by analyzing the structure of LiDAR and the physical principles during laser pulse propagation, as well as the mapping relationship between the properties of scenario elements and the risk factors was designed. The testing environment including the rainfall simulation system, diffuse reflective plates with different reflectivity, and vehicle targets was established. Static and dynamic validation tests were conducted to evaluate the impact of risk factors such as edge attenuation effect, target reflectivity, detection distance, and speed of LiDAR in different rainfall rates, and targeted improvement directions were proposed. Through verification, it can be seen that this method has a positive impact on improving LiDAR products and optimizing the weight allocation of ADS sensing sensor.