Research On Estimation Method Of Road Friction Coefficient Based On MEA Optimization BP Neural Network

With the rapid development of people’s living standards and transportation,as an important carrier of China’s economic development,the number of vehicles is also increasing.When the vehicle in the poor adhesion conditions of road surface such as water,ice,caused by a lack of drivers on the road to judge,easy to cause the instability of a vehicle and lead to traffic accidents.In order to ensure safe driving and reduce the occurrence of traffic accidents,it is inevitable to improve the active safety of vehicles.The road friction coefficient can directly reflect the road information and provide an important input parameter for the vehicle active safety system.This paper is devoted to improving the accuracy of estimating the road friction coefficient.A method of estimating the road friction coefficient by optimizing BP neural network with Mind Evolutionary Algorithm(MEA)is proposed,and the selection of dynamic response when estimating the road friction coefficient is studied.The main research contents of this paper are as follows.1.The CarSim/Simulink co-simulation model was established,and the simulation was carried out according to the designed sampling space.A total of 28 dynamic response data of 64 driving conditions of a vehicle were obtained,and the collected data were normalized,which provided the research basis for the following.2.According to the collected simulation data,the MEA-BP neural network model for estimating the road friction coefficient was built,and compared with the Extreme Learning Machine(ELM)and BP neural network.The results show that compared with ELM and BP neural network,the accuracy of MEA-BP neural network is increased by 8.9%and 5.7%respectively,and the mean-square error is decreased by 2.9E-03 and 1.5E-03 respectively,which has higher accuracy and stronger robustness.3.A vehicle was selected to install experimental equipment,and relevant experimental data were collected.Experiments on dry asphalt road and wet asphalt road were carried out to verify the effectiveness and feasibility of the method.4.According to the correlation analysis between the 28 dynamic responses collected by the CarSim/Simulink co-simulation and the accuracy of the estimated road friction coefficient,the optimal number of input variables to estimate the road friction coefficient and the optimal dynamics response were determined.The above studies on the estimation of road friction coefficient will improve and supplement the existing estimation methods,and provide theoretical and technical reserves for the study on the estimation of road friction coefficient.In addition,using MEA-BP neural network to estimate the road friction coefficient solves the defects of the current estimation method to a certain extent,and has important practical significance for improving the active safety of vehicles.