Risk Assessment Analysis Under Mixed Traffic Flow Scenario

Mixed traffic flow has gradually become a developing trend, in this trend, safety is the cornerstone of autonomous and manual driving vehicles. How to reduce the driving-related collision and improve people's feeling of road traffic safety is an urgent problem to be solved. In order to accurately identify and predict high risks and reduce security dangerous, we propose scenario-driven machine learning algorithms. Firstly, the Krauss model and the Intelligent Driver Model (IDM) are selected as the manual and autonomous driving models, four types of highway mixed traffic scenarios are reconstructed based on SUMO. Secondly, based on the collected simulation data, two type of risk identification algorithms are applied by using the logistic model and BP neural network model. Finally, the simulation results demonstrate that compared with the Logistic model, the BP neural network model has higher recognition accuracy and is more suitable for high-risk recognition and prediction.