Construction of Driving Hazard Disposal Model for Human Drivers Based on Natural Driving Data and Extraction of Critical Scenarios for Cut-out Scenarios

In the field of intelligent driving vehicle testing and evaluation, in order to achieve the coverage of scenarios, there are too many test scenarios required, and it is difficult to carry out all tests, and relevant standards and regulations lack specific parameters based on theoretical basis, resulting in insufficient credibility and persuasion. Based on the natural driving data of China-FOT project and the relevant provisions of ECE R157 regulations, this paper classifieds and extracts the natural driving data. In this paper, the type of front car cut-out with fewer scenes but high analytical value is selected for research, and the ECE R157 human driver emergency response model is introduced. It embodies the ultimate collision avoidance level of human drivers. Then, considering the driving experience of human drivers, this paper proposes a driving hazard disposal model for human drivers, and divides the scene collision avoidance difficulty into four levels according to deceleration gradient and maximum deceleration, which allows for a graded assessment of the ability of the intelligent driving car to cope with cut-out scenarios. Next, this paper uses the provisions of the relevant parameter values of the human driver emergency response model in ECE R157 regulations to extract 6 cases of test conditions from the 18,000 cases of logical scenarios, and the total test efficiency is increased by 3000 times. Then, according to the established human driver driving hazard disposal model, this paper obtains the boundaries of each difficulty scenario through simulation, which intuitively shows the scenario difficulty relationship between different scenarios and parameter values, and the critical scenarios of whether collision avoidance can be achieved or not. Finally, the paper analyses the distribution range of THW, a risk indicator for ordinary and excellent human drivers, under different scenario difficulties.