Research on Driving Style Recognition Model Considering Different Traffic Density and Driving Behavior

Aiming at the problems of single result and poor adaptability of existing driving style recognition models, a driving style recognition model considering different traffic densities and driving behaviors was proposed based on the influence of traffic density and driving behavior on driving style recognition. Using Prescan, Matlab/Simulink, and NIGSIM dataset to build a driving simulation platform to collect driving data of drivers on straight roads with different traffic densities. According to the collected data, the driving behaviors were divided and the feature parameters corresponding to the driving behavior were extracted. PCA analysis was used to perform post-processing of feature parameters, and based on the K-means++ algorithm, different driving behaviors under different traffic densities were clustered into three categories: conservative, normal, and aggressive. Based on the clustering results, the driving style recognition model was established by using a random forest algorithm, and compared with the driving style recognition model without subdividing traffic density and driving behavior. The result showed that the proposed driving style recognition model not only had high recognition accuracy but derived the driver's comprehensive driving style. At the same time, it can improve the adaptability of the recognition model to different traffic environments and driving behaviors.