Driving Style Recognition Incorporating Risk Surrogate by Support Vector Machine

Paper in proceeding, 2021

Accurate driving style recognition is a crucial component for advanced driver assistance systems and vehicle control systems to reduce potential rear-end collision risk. This study aims to develop a driving style recognition method incorporating matching learning algorithms and vehicle trajectory data. A risk surrogate, Modified Margin to Collision (MMTC), is proposed to evaluate the collision risk level of each driver’s trajectory. Particularly, the traffic level is considered when labelling the driving style, while it has a great impact on driving preference. Afterwards, each driver’s driving style is labelled based on their collision risk level using the K-means algorithm. Driving behavior features, including acceleration, relative speed, and relative distance, are extracted from vehicle trajectory and processed by time-sequence analysis. Finally, Supporting Vector Machine (SVM) is applied to recognize driving style based on the extracted features and labelled data. The performance of Random Forest (RF), K-Nearest Neighbor (KNN), and Multi-Layer Perceptron (MLP) are also compared with SVM. The “leave-one-out” method is used to validate the performance and effectiveness of the proposed model. The results show that SVM over performs others with 91.7% accuracy. This recognition model could effectively recognize the aggressive driving style, which can better support ADAS.