| As an important part of the traffic flow theory,the merging behavior not only needs to consider the running state of the merging vehicles and surrounding vehicles,but also the interaction between the main line leader and the merging vehicles.Therefore,the merging behavior is relatively complex and the risk is higher.At this time,the vehicle merging assistance system,as an important part of the driving assistance system,can provide the driver with safe merging decision suggestions to reduce the risk of collision.Since vehicle merging generally takes several seconds to execute,in the process of merging execution,there are complex interactions between the merging vehicle and multiple vehicles on the original lane and the target lane.The analysis and modeling of deceleration behavior can provide theoretical support for the development of safer and more efficient driver assistance systems.In order to study the interaction in the process of convergence and interaction of traffic flow in the expressway weaving area,this paper firstly determines the research scene and object,qualitatively analyzes the convergence and interaction behavior of the expressway weaving area,reveals the internal operation mechanism of the convergence and interaction behavior,and adopts the gradient boosting decision tree.and multivariate adaptive regression splines to build the intersection interaction model in the weaving area respectively,and introduce the speed difference,distance difference,lateral position and safety influencing factors between the main line and the following and merging vehicles and the surrounding traffic flow to analyze the convergence of traffic flow in the weaving area.Interactive behavior;secondly,use the US-101 vehicle trajectory data in the American Next Generation Simulation(NGSIM)data set to establish data cleaning rules,extract relevant influencing variables for model training and testing,obtain model operating parameters,and determine the importance of influencing variables percentage,and compare the fitting effects of different models on the merging vehicle and the main-line collar-following vehicle,and predict the longitudinal acceleration of the merging vehicle and the main-line collar-following vehicle.The validity of the proposed model is verified from the perspectives of flow operation efficiency and traffic flow safety.The research results show that under the interaction of different influencing variables of interactive traffic flow,the prediction accuracy of the merging vehicle is significantly higher than that of the main-line leader following.The prediction accuracy of the adaptive regression spline model is much higher than that of the perspective-based stimulus-response model,and slightly lower than that of the gradient boosting tree model,but the complexity of the multivariate adaptive regression spline model is much lower than that of the gradient boosting decision tree,and can Providing corresponding explicit expressions can effectively reveal important data patterns and relationships,and reflect the interaction between different influencing variables,which is beneficial to the application in assisted driving systems and unmanned driving systems,and can also help engineers and researchers.Gain an in-depth understanding of the different interaction rules in lane changing behavior.The study found that both the gradient boosting decision tree and the multivariate adaptive regression spline model can effectively predict the interaction behavior of the traffic flow in the weaving area.Variable transformation,dealing with complex structures hidden in high-dimensional data,the accuracy of the prediction results is relatively high,and the obtained vehicle longitudinal acceleration curve is also relatively smooth,and the model can be fitted for different data sets.Predicting performance is beneficial for applications in assisted driving systems and unmanned systems. |
