| About 80% of traffic accidents on the freeway occur at the interchange entrance and exit areas,and the intertwining formed during the single mandatory lane change at the entrance and exit and insufficient deceleration adjustment into the ramp are the main causes of accidents at the interchange exit.In the previous studies,as an important part of the lane change process,lane change decision-making has received much attention and rich research,but the depth of research on the subsequent execution stage and the post-adjustment stage is relatively lacking.From a micro perspective,this paper divides the single lane change process into three stages: decision-making,execution and adjustment,and mainly focuses on the execution and adjustment stages of a single mandatory lane change for vehicles at the entrance and exit.The research results can provide a reference for the improvement of the micro lane change model and the macro lane change distance calculation mode.Firstly,US101 section of the NGSIM dataset is selected as the lane change data source,while database was emplyed to extract the entry and exit lane change vehicle data.Subsequently,wavelet is employed to denoise and reconstruct the vehicle trajectory data that will be used for following analysis.Secondly,for the implementation phase of a single mandatory lane change at the entrance and exit:(1)According to the definition of the execution stage,combined the vehicle’s operating characteristics with traffic environment characteristics,more than ten indicators for the lane change are extracted as the influencing factors of the execution time and then analysed;(2)Multiple regression,BP neural network and improved BP neural network based on genetic algorithm are separately emplyed to fit and predict the entrance and exit execution time.The results show that multiple regression can extract the main influencing factors and quantify the impact of each factor on the lane change execution time,but the prediction effect is not as good as the BP neural network.The improved BP neural network with the genetic algorithms operation faster and predict better;(3)Based on the above research,the characteristics of the longitudinal distance of the lane change execution were investigated,and it was found that the initial speed and execution time of the lane change were the main factors affecting the lane change execution distance.The nonlinear regression and multiple linear regression were established with the two factors as independent variables,which shows a good fitness and prediction effects.Then,for the adjustment phase of a single mandatory lane change at the entrance and exit:(1)Based the separation the adjustment phase,considering the vehicle’s operating characteristics and the relationship between the car groups,the first and second judgments criterion of the lane change adjustment end point are subsequently proposed,and next correlation analysis of the factors affecting the lane change adjustment time are done,from the view of lane change vehicle and the traffic environment;(2)Multiple regression and neural network are employed to fit and predict the entrance and exit adjustment time,and the results show that the neural network performs better in fittness the prediction effects.Besides,the neural network operation efficiency can be improved after introducing the genetic algorithm;(3)The relationship between the longitudinal distance of the entrance and the lane adjustment time and speed is explored.Based on that,a regression equation is established and behaves well in prediction.Finally,focused on the type A interleaved entrance and exit lanes,based on the study of time and longitudinal distance during the lane change execution and lane adjustment phase,the weaving section length that meet the lane change needs is recommended from the lane change perspective.Besides,various region specifications and standards related to weaving section length are compared. |
PDF Full Text Request