Modelling of Longitudinal and Lateral Behavior of Drivers and Automatic Prediction-evaluation

The rapid development of intelligent vehicles has also advanced the capabilities of the Advanced Driving Assistance System (ADAS) in mass-produced vehicles. ADAS is essential to reduce the workload of human drivers and enhance their safety. However, irrational and unexpected driving behavior, which may occur, especially in high-stress scenarios, can lead to vehicle instability and the ADAS may be unable to assist sufficiently. This is a situation, where it would be desirable to anticipate the drivers' behavior, so that it can be integrated in the dynamic model of the vehicle. The anticipated driver action could be regarded as the system input, which is required to predict the vehicle state and improve the performance of the vehicle controller. In this study, we use a data-driven approach to predict a drivers' longitudinal and lateral behavior, simultaneously and online. The study also compares and selects different input-dimension combinations of driver models to demonstrate its correlation with vehicle motion. Finally, the concept of 'prediction confidence' is introduced, which characterizes the accuracy of the prediction results. This parameter can be viewed as a criterion to control the vehicle system adequately.