Vehicle Collision Avoidance Algorithm Of Criticality Safety Vehicle-vehicle Distance In Vehicle Active Safety

In the past20years the rapid growth and the popularity of passenger vehiclesmakes China a superpower market for vehicle sales. The country is also facing severeproblems with vehicle traffic safety due to a large number of traffic accidents andcasualties. With the development and progress of social economy and the improvementof people’s living standards, higher demands are made for traditional vehicle safetytechnology. With the further development of passive safety techniques, at the sametime increased attention is focused on vehicle active safety techniques.This thesis aims at the development of an online intelligence system in which amathematical model of the vehicle critical following distance is combined with BPneural network technique. The system is used during the analysis of vehicle safetydistance and early warning of the driver to reduce the incidence of accidents. After astudy of the existing techniques for the vehicle active collision warning/avoidancecollision (CW/CA), it was found that most mathematical models of criticalfollowing distance lack consideration for the variables regarding driver individualcharacteristics in the actual driving process. Thus, a BP neural network model wasestablished by using the following vehicle’s initial velocity (v1) and the front of thevehicle’s initial velocity (v2) as the input variables. The parameters reflected the driverindividual characteristics in the driving process, such as the relative deceleration (a),and the eventually following distance (sfl) are defined as output variables. Thecalculations of the relative deceleration (a) and the ultimately following distance (sfl)were carried out using the BP neural network model. And then the model of BP neuralnetwork was implemented by MATLAB for feasibility study and simulation analysis.The values from BP neural network predictions are compared with values from theactual driving process by using characteristics of error back-propagation in the BPneural network. Through adjustment of the network weights and threshold in real time,the BP neural network predictive value is closer to the value in the actual drivingprocess. The final distance was used as a warning signal according to the mathematicalmodel of safety distances, in order to ensure access to distance with numericalaccuracy.Finally, a prototype system is designed and developed using the corresponding C# program, which was implemented to the whole car-following process and simulationanalysis, in order to conform to a more realistic environment adapting to the drivingbehavior, and to improve the prediction accuracy of the alarm for the critical distancebetween running vehicles, and further reduce the errors, such as reports of falsepositives or false negatives. In summary, this study provides a theoretical basis and apractical application for the design of a CW/CA system.