Research On Braking Distance Of Vehicles Prediction Based On Artificial Neural Networks

Driving safety related to people's lives and property, automobile brake performance directly related to driving safety, and vehicle braking distance is one of the most important indexes that measure vehicle braking performance. Therefore, studying vehicle braking distance has great significance. Although some scholars achieved fruitful results on the research of vehicle braking performance and braking distance, research on the automobile braking distance prediction was relatively smaller. Research on braking distance of vehicle based on artificial neural networks provide references for vehicle collision avoidance system,brake warning and so on. The main content of this research and the results as follows:(1) Based on the analysis of braking process, influence factors of braking distance and adhesion coefficients, the calculation methods of braking distance, and the elaboration of the prediction theory of artificial neural networks, the thread of automobile braking distance prediction by using artificial neural networks was proposed, and the necessity and feasibility of automobile braking distance prediction were elaborated.(2) The thread of establish of braking distance prediction model based on artificial neural networks was proposed by using parameters of vehicle driving conditions as input and output samples, getting the network weights and threshold value of each layer, and then getting solidified prediction models by training the neural network. The vehicle braking distance models were established and simulated by using MATLAB software based on four type road conditions, which included dry asphalt and concrete roadway, dry gravel roadway, wet asphalt and concrete roadway and ice-snow roadway. The result showed: the average relative error between braking distance of the four models estimated value and sample value were:-0.13%, -0.76%, 1.04% and 0.33%.(3) The vehicle braking distance prediction system's software was designed and finished. And the results of the test experiment of the prediction system showed that the system could calculate the adhesion coefficients and parameters of braking distance; predict vehicle braking distance with manual and automatic ways. This system could accept velocity and adhesion coefficients from peripheral equipment, display these parameters and braking distance at once and store them in ACCESS database. When vehicle velocity was between 30km/h and 120km/h, the average relative error of forecasting braking distance was 0.99%, among them the largest relative error was -3.18%, the least relative error was 0.66%.The research results showed that the prediction way of using artificial neural network to predict vehicle braking distance was feasible, the system running was dependable, and the prediction precision was high.