Modeling For The Autonomous Merging Strategy Of Intelligent Vehicles Considering The Remaining Length Of The Acceleration Lane

Autonomous driving has been a hot topic in the two fields of market and scientific research in recent years.In the short term,the mixed mode of autonomous vehicles and traditional vehicles will continue to exist.During the merging process of autonomous vehicles in the acceleration lane of the expressway,the probability of traditional vehicle drivers actively avoiding autonomous vehicles is not high,and the import is different from the lane changes,autonomous vehicles must be merged within the remaining length of the limited acceleration lane,which leads to the need for autonomous vehicles to be fast and accurate in the process of decision-making.Aiming at the above-mentioned problem,the ESR radar and the driving recorder are installed on the inbound section of an expressway hiddenly.In the case of free traffic on the National Day,a large amount of imported data was continuously collected.Analyzing the imported data of 619 groups of vehicles randomly,the import scenes are divided into 4 types according to whether there are front and rear cars in the import process.Based on the analysis of the import duration,line pressing time,the length of the remaining acceleration lane,vehicle speed,whether the turn signal is turned on,etc.the driver’s behavior parameters are acquired.Then this article focuses on the establishment of a random forest model for the latter two scenes,and this model is applied to the decision-making process of the intelligent vehicles.The main contents and conclusions of the research are as follows:(1)The duration and pressing time of the latter two types of scenes are greater than those of the first two types of scenes;Regarding the remaining length of the acceleration lane,in the first type of scene,the driver is more likely to realize the merging farther from the end of the acceleration lane,and in the third and fourth types of scenes,the driver’s merging position is closer to the end of the lane;Regarding the relationship between the line pressing time and the average speed of the vehicle,In the second types of scenes,the line pressing time varies slightly with the speed of vehicle.The other three types of scenes have larger correlation coefficients;Regarding whether the turn signal is turned on,the fourth type of scene is the largest,and the first type of scene is the smallest.(2)In the third type of scene,the decision efficiency reaches 88.89%,in which the relative distance,the relative speed between two vehicles,and the remaining length of the acceleration lane are used as input,the mode of the waiting for import and direct import are used as output.the mode of direct import is further subdivided into the mode of safety and risk.The overall decision-making efficiency can reach 83.33%,in which the three-parameter are used as input and three-parameter are used as output,which can still meet the requirements of the model.(3)In the fourth type of scene,the overall decision-making efficiency of the combination reaches 87.50%,in which the relative distance between two vehicles,the relative speed of the vehicle and the driving environment are used as input,and the output is the mode of waiting for import and direct import.Finally,the decision model was applied to Matlab to establish the corresponding conditions of simulation,which verified the feasibility of the latter two scene.(4)The decision-making efficiency of the combined model with remaining length of the acceleration lane parameters is better than the combined model without remaining length of the acceleration lane in the third and fourth types of scene,and the combination that considering the remaining length of the acceleration lane achieves the highest decision-making efficiency in a shorter time window,which also proves the importance of the remaining length of the acceleration lane.Also,it takes a longer time to consider the safety of the import decision than not to consider whether the import decision is safe.