Study On The Car-following Behavior Based On Multi-information Fusion

The driving behavior model can be used to recur actual dynamic behavior of traffic flow under all kinds of roads and transportation condition. It can reflect random characteristic of running vehicle and individual preference of the different driver groups in the car-following and lane-changing process. So, research of driving behavior has become the theoretical basis and pivotal tache about transportation system simulation research and an important means to exploit and study on traffic field, especially for ITS which has been widely carried on. The characteristic of the multi-information stimulation, the task concentration of the driver, and the cooperative reaction behavior have not been considered synthetically in the traditional driving behavior model which was mostly centered on car-following model. In present situation of domestic research field, multi-information fusion has already been widely studied and applied in electronic and military domain, and have been used in traffic incident detection and traffic flow forecasting of traffic engineering domain. However, it has not been made the best of in studying driving behavior.In this paper, the driving behavior on the single lane is carried out to study at large from theory lay. The psycho-physical integrated cognitive topological structure of driving task-centralization under the multi-resource information stimulation is designed. The multi-information fusion algorithm based on fuzzy integral theory and Bayes theory is built. At the same time, the fuzzy integral fusion algorithm is used to analyze and get the running pattern of the vehicle in the complex running environment. The car-following model based on synchronizing movement between two vehicles is presented and the car-following model is combined with the multi-information fusion algorithm to confirm the driving behavior. At last, the non-parametric method is used to demarcate and validate the model and algorithms through concrete data. The results show that the model and algorithm are effective.