Driving Behavior Recognition And Optimization Based On In-vehicle States

Driving behavior affects economic and safe driving very much. Reasonable assistance and optimization of driving behavior can reduce fuel consumption and increase equality of driving safety. Thus, this paper put forward the methodology of driving behavior recognition method based on the in-vehicle state and optimal strategy of driving behavior. Data base for driving behavior recognition is provided by building a distributed in-vehicle network and digital collection of vehicle state information. The method of driving behavior recognition is established on the basis of driving data. Through statistical analysis of fuel economy of different driving laws, combining with economic velocity based on dynamic vehicle mass, we provide optimal recommendations for vehicle economy.The network sharing and digital collection of vehicle state is achieved by proposal of distributed in-vehicle network architecture of independent coordination. By designing vehicle state information, developing in-vehicle bus protocol mechanism and establishing the coordination function of vehicle state to achieve network control. Proposing vehicle network node optimal problem and solving optimal network node through backtracking algorithms. Optimizing harness of network, automotive wiring harness by mapping the surface, establishing harness optimization constraint equations, using gradient descent method to solve optimization problems of harness.An improved theory of driving behavior recognition of excellent performance of higher accuracy and computational efficiency based on dynamic time window and Hidden Markov Models is established by the proposal of driving behavior model. Vehicle steering and lane changing behavior recognition method based on Supported Vector Machine is built by extracting diverse steering and lane changing behavior feature on the basis of vehicle steering kinematics.Vehicle fuel consumption monitoring is achieved based on building fuel consumption calculation model by utilizing bus data from in-vehicle network whose parameters calibrated via fuel experiment. We Propose an economic evaluation system of driving behavior economy and excavate economic driving behavior laws through statistical analysis of experimental data. Proposed optimal strategy based on economy velocity is built by vehicle longitudinal dynamics based on dynamic vehicle mass model adapts to driving condition and road condition.Vehicle test platform is built by transformation of the real car which can be utilized to verify the in-vehicle network. Driving behavior recognition and optimization are tested by analysis of experimental data.