Study On Feature Extraction And Identification Model Of Driving Tendency Based On Dynamic Driver-Vehicle-Environment Data

With the rapid development of the economy,there is a significant increase in number of car ownerships,especially private cars.The contradiction of human-vehicle-environment is serious in road traffic system.Above 90%factors of traffic accidents is human factor and above 70%traffic accidents are caused by drivers.Driver’s physiological and psychological characteristics are closely related to traffic safety and its influences on traffic safety are mainly represented as driver’s tendency.Most of the used researches focused on the influence on traffic safety and the driver ’s psychological characteristics from the relative static and macroscopic perspective.However,in the fields of vehicle active safety,there are few researches on driver’s affective characteristics which are measured and calculated from microcosmic and dynamic perspective.Micro-dynamic information such as vehicle state,driver’s behavior,driving environment can be obtained in designed various experiments.The data gained are used to training the model of neural network classifier and the back-propagation algorithms in order to ascertain connection weights between input layer,output layer and the hidden layer,to get the correct classification rate for estimation during the process micro-dynamic information is regarded as the neural network input layer and the driver’s preference types are regarded as the output layer.At last,the data of driver’s tendency can be extracted by the Discrete Particle Swarm Optimization Method.Two-lane road is taken for example and state factors which directly impact driver’s affection are paid important attention.Characteristics of driver’s tendency toward different environments are extracted using genetic simulated annealing algorithm.Identification model of driver’s tendency which is adaptable to changing environment is established by dynamic Bayesian network.Under the two-lane road condition,the metastatic patterns of driver’s tendency as the environment changes is revealed through analyzing the data of driver’s tendency in different environment.Results show that the established identification models are adaptable to realize the real-time identification of driver’s tendency type in multi-lanes road.It also can provide theoretical basis for the realization of personalized automobile active safety system.