Traffic Flow Steering At Intersections Based On Trajectory Data Research On Traffic Prediction Method

With the rapid development of road traffic,the problem of urban road traffic congestion has become increasingly prominent.Intersection congestion has become the pain point of traffic management.Traffic signal optimization is one of the effective ways to solve the congestion of Road intersection.However,for some remote intersections and intersections with missing detection equipment,signal optimization can not achieve the optimal control effect due to the lack of accurate traffic turn flow data.However,with the increase of the number of floating vehicles in the road network and the accumulation of trajectory data in the process of vehicle operation,it is possible to predict the turning flow at intersections through trajectory data.In this paper,aiming at the problem of the lack of traffic turn flow data in the remote and lack of detection equipment,in the aspect of data preprocessing,the data is first matched with the map.Then the data of different flow directions are selected according to the changes of the points before and after the data.Then,an intersection flow calculation model based on traffic wave theory is designed,and the actual traffic flow of the road is calculated by using the floating vehicle data.Due to the strong instability of traffic data,there are a large number of high-frequency noise data in the data,and the traffic data has a certain dependence on the time dimension.Therefore,a short-term combination forecasting model of turn traffic flow based on empirical mode decomposition(EMD)and gate recurrent unit(Gru)is proposed.Empirical mode decomposition(EMD)is used to denoise the calculated flow data.This method can effectively reduce the impact of noise data on the prediction accuracy.Then the decomposed modal component data is input into the model according to different flow directions for prediction.In the prediction process,the historical flow data is used as the input feature to predict the results.Based on the track data of taxi driving for 10 days in Zhongguancun area of Beijing,the algorithm is verified by dividing the data into three periods: morning peak,flat peak and evening peak,and compared with LSTM model and Kalman filtering algorithm,which are widely used.The algorithm proposed in this paper has higher prediction accuracy and faster prediction speed.The model can be applied to short-term traffic flow turn prediction at intersections,and can provide accurate turn flow data for remote or missing intersections.In addition,the predicted data can complement the traffic data of the intersection with detection equipment to get more comprehensive traffic data,which provides good data support for the subsequent signal optimization of the intersection.