Fast Decision-making Method For Personalized Lane Change In Highway Merging Area Of Vehicle Based On ELM

Lane change in merging areas,entrances of highway,is very complicated.Due to frequent speed varying and lane change in acceleration lanes and main trunk traffic roads,merging areas are of high-incidence for traffic safety accidents,in which faster and more accurate decision-making is required.Therefore,decision-making methods are important to safe driving.When drivers change lanes in merging areas,the current personalized lane change model in the Advanced Driver Assistance System(ADAS)is too complicated,and often combines multiple algorithms for calculations,making the computation of the lane change model more complicated.To improve the comfort level for drivers on the premise of safe and stable driving,this paper designs a decision-making model for personalized lane change according to lane change behaviors in merging areas of highway and driving habits of different drivers.This paper has completed the following work:It confirms the characteristic quantity of representative personalized lane change based on the analysis of the relevant driving characteristics of vehicles in merging areas of highway and relevant influential elements of lane change.On the basis of actually measured vehicle dataset of Next Generation Simulation(NGSIM),it draws a trajectory map of the vehicle by means of MATLAB,and gives the data in the lane change stage and the non-change stage of the mandatory lane Change by using trajectory change trend.According to the different driving habits of different drivers,based on the Extreme Learning Machine(ELM)and online learning method,the model training is divided into two stages: offline and online.The personalization of the lane change decision-making model is achieved through training according to online drivers' driving data.At the same time,it solves the problem that the new model needs to be retrained when a new sample is added,which saves the running time of the model.Based on Online Sequential Extreme Learning Machine(OS-ELM),a fast decision-making model for personalized lane change in merging areas of highway is constructed.According to the dataset of different roads,a simulation experiment is made on the test precision and test time of lane change stage and non-change stage of the decision-making model on the basis of OS-ELM,SVM and BP.With this experiment,it tests the effectiveness of the decision-making model for lane change created by this paper.The OS-ELM-based personalized lane change decision-making model proposed in this paper meets the speedability in merging areas of highway and gives time for drivers to make decisions.The proposed model provides a reference for the improving design of ADAS core components,so that a better ADAS system will be developed.