Research On Short-term Traffic Flow Forecast And Signal Control Based On Deep Learning

This article first introduced the concept of deep learning,and then selected several deep learning models and improved them to adapt to short-term traffic flow prediction.There are three main models: an improved LSTM model,a model combining spatio-temporal analysis and GRU,and CNN-LSTM model.Through the use of actual traffic flow data,these three models were used for learning and training,and finally the three deep learning models were tested with data.Through comparative experiments,the improved LSTM model is selected as the short-term traffic flow prediction model in this article.On the basis of short-term traffic flow forecasting,this paper establishes a single-intersection multi-objective optimization model targeting average vehicle delay,intersection capacity and CO emissions,and solves it with an improved particle swarm algorithm.The actual case and the status quo are compared and analyzed,which proves the effectiveness of the model.After completing traffic flow prediction and single intersection control,in order to achieve the final coordinated control of regional traffic,it is necessary to divide the control area into traffic sub-areas.This paper uses the SAEs-LSTM model to dynamically divide the target area.The real-time intersection signal cycle,the flow rate and distance between adjacent intersections are used as the input of the SAEs-LSTM model,and the demand intensity coordination coefficient for coordinated control between adjacent intersections is predicted and output,and the traffic control sub-system can be realized according to the real-time characteristics.Dynamic division of zones.Finally,threshold control is performed on the road network selected in this article,the threshold value is determined according to the historical flow of the area,and the traffic flow outside the core road network is optimized by signal lights to induce and divert,and the traffic flow into the core road network is controlled In the unsaturated state;then,for the core road network,the traffic flow after threshold control is subjected to local regional traffic coordination control.By selecting the key intersections and key trunk lines of the road network,the green wave coordination method of the trunk lines is used on the key trunk lines,and other intersections adopt the aforementioned single-intersection adaptive control method to realize the coordinated control of regional traffic.