Study On Strategy And Control Method Of Intelligent Vehicle In Highway Meriging Area Under Mixed Traffic Situation

The total quantity of vehicles in our country has increased year by year,but the issues,such as traffic congestion,environmental pollution,are also becoming serious.Countries are vigorously developing intelligent driving vehicles,hoping to improve traffic efficiency and reduce pollution emissions by intelligent strategy.Freeway on-ramp merging area is one of the special traffic scenarios,the merging behavior of on-ramp vehicles greatly affects the traffic efficiency.Therefore,this paper focuses on the problem of decision control in the on-ramp merging area of the freeway,and establishes the lane change decision model and the car-following model of the mixed vehicle fleet by using the improved reinforcement learning algorithm.The main innovations and contributions of this paper are as follows:(1)The interaction characteristics of vehicles in the on-ramp merging area of the freeway were analyzed,and the mean road space-time utilization rate model was proposed.Considering the spatial distribution and time distribution,the carrying capacity of the road was maximized,so as to improve the traffic efficiency of vehicles in the on-ramp merging area.(2)The hyperbolic cosine Q value estimator was innovated.Three simple neural networks are used to calculate the coefficients of the hyperbolic cosine function,and the final Q value is obtained by calculating the hyperbolic cosine function.This structure can transform the complex neural network into several simple neural networks,reduced the computation time and improved the timeliness of decision.The hyperbolic cosine Q estimator was applied to the deep Q learning algorithm,and the lane change decision model of intelligent driving vehicle was established.(3)According to Mirror-Traffic data set,the K-means clustering algorithm was used to classify the driving styles of drivers in the data set based on the three parameters of maximum speed,acceleration and time headway.In the SUMO simulation software,the traffic scenario of the on-ramp merging area of the freeway was built,and the related motion parameters of the vehicles in the traffic flow were set up according to the clustering results,so as to establish the mixed traffic flow of human driving vehicles and intelligent driving vehicles.The simulation results show that under three different traffic flow conditions,the traffic efficiency increases by33.3%,26.3% and 20.5% respectively,and the fuel consumption of vehicles decreases by 7.35%,8.1% and 7.74%,respectively.Accordingly,vehicle emissions of carbon monoxide,carbon dioxide and other pollutants also fall.(4)Aiming at the research on the control of mixed traffic of the freeway on-ramp merging area,the car-following model of mixed traffic fleet based on hyperbolic cosine DDPG algorithm was proposed.In the SUMO simulation software,the car-following scene of whole human driving and of mixed human driving and intelligent driving were set up.They were taken as the comparison test groups,and the speed,acceleration,fuel consumption and pollutant emission of vehicles in the fleet were analyzed statistically.The comparison results show that the variance and standard deviation of the vehicle speed and acceleration are reduced,and the hyperbolic cosine DDPG hybrid car-following model can effectively attenuate the speed oscillation wave of the pilot vehicle and improve the stability of the vehicle string.At the same time,vehicle fuel consumption decreases by about 15.3 percent,carbon monoxide and carbon dioxide emissions decreases by5.34 g and 280.54 g.