Research On Variable Lane Clearance Control Based On Traffic Flow Prediction

Tidal traffic is a common traffic phenomenon in many cities.The unbalanced flow of tidal traffic is likely to cause traffic congestion and insufficient road resources on one side,and sparse traffic flow and inefficient utilization of road resources on the other side.Therefore,variable lane control can provide reasonable measures to alleviate the tide traffic congestion and improve the utilization rate of road space-time resources.Based on the above problems,this paper first conducts research on the traffic flow prediction problem.Considering that the existing short-term traffic flow prediction methods have shortcomings in the study of the temporal and spatial characteristics of traffic flow,which leads to the low prediction accuracy,this paper proposes the Pearson-Dynamic Time Warping Distance(PD)-Graph Convolutional Neural Network(GCN)-Long Short Term Memory Network(LSTM)short-term traffic flow prediction model.On this basis,a variable lane control model based on short-term traffic flow prediction is constructed,combined with variable lane clearing time and upstream and downstream intersection signal control strategies to determine the optimal switching time of variable lanes to achieve the effect of rapid emptying of the variable lane.The main research contents of this article are as follows:(1)Spatio-temporal correlation analysis method based on traffic flow time series.Aiming at the problem that the variation trend and distance characteristics of traffic flow time series require quantitative analysis,PD distance is proposed to measure the space-time correlation degree of traffic flow time series.It is a comprehensive correlation analysis method that uses pearson correlation coefficient to conduct trend quantitative analysis of traffic flow time series and dynamic time warping distance to conduct distance quantitative analysis.(2)A short-term traffic flow prediction model based on tempora l and spatial correlation.Based on the analysis results of the temporal and spatial correlation of traffic flow time series,the threshold of PD distance is set to realize the screening of related sections in the road network,and further construct the set of temporal and spatial correlation sections,and establish the PD-GCN-LSTM traffic flow prediction model.Finally,an example of traffic flow prediction is analyzed on a selected road segment in the road network.By comparing the prediction results of LSTM model,GCN model,PD-LSTM model,PD-GCN model,GCN-LSTM model and PD-GCN-LSTM model,the effectiveness of the proposed PD-GCN-LSTM traffic flow prediction model are illustrated.(3)Coordinated optimization method for clearing variable lanes and signal timing of upstream and downstream intersections.Aiming at the conversion condition of variable lane function,a variable lane control model based on short-term traffic flow prediction is constructed.On this basis,the variable lane segment clearing time and the upstream and downstream signal control strategies in different states are studied,and a coordinated control model of variable lane clearing and upstream and d ownstream intersection signal timing is established to obtain the best opportunity for variable lane function switching.