Research On The Vehicle Collision Warning/Avoidance Model Considering The Intention Of Front Vehicle Driver In The Internet Of Vehicles Environment

With the rapid increase of car ownership in China,traffic accidents occur frequently and traffic safety problems have become increasingly prominent,among which rear-end collisions are the most common.Rear-end accidents usually occur suddenly.If the rear vehicle can obtain the intention information of the driver in front of the vehicle in time,it can remind the driver of the rear vehicle to brake or trigger automatic emergency braking to avoid collision when the driver of the front vehicle takes emergency operations.Therefore,this thesis relies on the support of National Key R&D Program of China: Development and Application of a New Type of Multi-functional Intelligent Vehicle Terminal(2018YFB1600701)and takes the driver’s intention recognition in front of the vehicle in the Internet of Vehicles environment and the corresponding rear vehicle collision warning/avoidance model as the research object.Then,the acquisition and selection of driver’s intention observation data,the driver’s intention recognition method and the calculation method of critical safety distance,etc.which are involved in the realization of the collision warning/avoidance function,were analyzed and researched.So,a vehicle collision warning/avoidance model considering the intention of front vehicle driver in the Internet of Vehicles environment is proposed.Firstly,this thesis builds a real vehicle acquisition platform for drivers’ intention observation data,including acquisition hardware and acquisition software.The platform is used to collect the original data of driving operation and vehicle state of different drivers under different intentions.After the preprocessing based on exponential moving average filter and the analysis of intention observation data,the brake and acceleration pedal displacement and their rate of change and vehicle speed were selected as intention characterization parameters.In order to provide reliable data support for the subsequent establishment and verification of the driver’s intention recognition model,a driver’s intention observation database with different time window parameters was constructed by extracting the intention sequence of the representational parameters in a sliding time window manner.Secondly,according to the characteristics of driver’s intention generation process and corresponding observation data,a double-layer recognition architecture is adopted for drivers’ intention.Back-Propagation Neural Network(BPNN)and multi-dimensional discrete Hidden Markov model(HMM)are selected as the main models of driving behavior layer and driving intention layer,and their key parameters are analyzed and selected,so as to construct the intention recognition model based on BP-HMM,and realize the recognition of accelerated driving,uniform driving,normal braking and emergency braking intention.Compared with the traditional recognition models based on BP or HMM,the intention recognition model based on BP-HMM has a higher recognition accuracy on the premise of guaranteeing real-time performance.Then,the factors affecting the calculation of critical safety distance such as driver intention recognition delay,network communication delay and road adhesion coefficient in the Internet of Vehicles environment are analyzed.Based on the vehicle braking process model,this thesis constructed the corresponding collision warning/avoidance critical safety distance for the five working conditions of the front vehicle driver,including accelerated driving,uniform driving,normal braking,emergency braking intention and communication failure,so that the rear vehicle can dynamically adjust the collision warning and automatic emergency braking logic according to different working conditions.Finally,in order to verify the feasibility and effectiveness of the proposed vehicle active collision warning/avoidance model considering the intention of the front vehicle driver in the Internet of Vehicles environment,this thesis built a driver-in-the-loop co-simulation platform based on Pre Scan,Carsim and Simulink,and a comparative simulation test of multiple working conditions and multiple models was carried out.Results show that the active collision warning/avoidance model proposed in this thesis is better than traditional TTC,Mazda,Honda,Berkeley,and vehicle collision avoidance time model based on driving speed,without interfering with the normal operation of the driver.High correct early warning rate,and can successfully avoid collisions in automatic emergency braking test conditions,with higher safety and stability.