Car-following safety modeling and risk assessment of autonomous vehicle in icy and snowy weather

This paper is to study the effect of icy and snowy weather on the car-following (CF) safety of autonomous vehicle (AV). The influence of weather is abstracted as mathematical model parameters, and the CF model and risk decision equation of AV under icy and snowy weather are constructed. Comparing the influence of various climates on the CF of AV and the potential safety hazards, the CF parameters of AV in icy and snowy weather are designed based on Intelligent Driver Model (IDM). The road friction coefficient is matched by the maximum acceleration and comfortable deceleration of the vehicle, and the perception error coefficient is identified by space headway and speed of the vehicle. The Waymo dataset is used as the basic data source, and the safe value interval of icy and snowy parameters is calculated by combining the CF equation and the dataset characteristics. The rationality and stability of the parameters are verified by the root mean square error (RMSE) method and the Wilson model. Using the SUMO platform, single and multiple factors scenes are designed for simulation experiments, and a safety field strength model is constructed to carry out CF risk assessment. It is found that the severity of icy and snowy weather significantly affects the road friction and perception error coefficient, and has strong safety disturbance to the driving speed and real-time headway of AV. The accelerated and decelerated CF will cause oscillation and change of autonomous driving traffic flow, and the fluctuation range and risk degree of the queue caused by decelerated CF is more pronounced than that caused by accelerated CF. The safety effects of CF vary with different icy and snowy coefficients, and the vehicle speed perception error is more likely to induce safety risks. This study further enriches CF methods in special scenes, providing the theoretical basis for AV winter travel.