Research On Smart Control Over Regional Traffic Signal Light Based On RFID Data

With the development of economy,the traffic congestion problem is highlighted.Intelligent control of regional traffic Signals of the intelligent transportation system can solve the problem of uneven distribution of the duration of passage and alleviating the regional traffic congestion.At present,the more common traffic signal control method is a fixed timing strategy,which sets the traffic light length according to the traffic volume of the intersection,and will no longer be changed once it is determined.The main problem caused by this method is that it can not dynamically adjust the traffic light length according to the change of traffic flow,resulting in uneven distribution of traffic in each direction.Real-time traffic flow detection is the difficulty of realizing intelligent control of regional traffic signal.The traditional video recognition method and the magnetic induction coil identification method are difficult to obtain the vehicle information,but the radio frequency identification(RFID)method can obtain the vehicle information and thereby identify the vehicle steering.This thesis presents a method to implement the intelligent control of regional traffic signal by RFID data of intersection vehicle.The method is divided into three steps.The first step to deal with the vehicle steering identification,which uses the method of vehicle phase identification based on time phase division to identify the vehicle steering of RFID data,and then statistics the traffic flow.The main purpose is to carry out data labeled.The next step is to construct the steering vehicle traffic forecast model by feature set and model training through the labeled data based on distributed machine learning algorithm.The last step is to complete the intelligent control of traffic signal.It uses the output of the steering flow prediction model as the basis for decision-making to implement the intelligent control of regional traffic signal using reinforcement learning algorithm.In this thesis,stochastic forest algorithm based on Spark is research to forecast traffic flow and Q-Learning reinforcement learning based on BP neural network is used in intelligent control of traffic signal.According to the results of the analysis of data and validation of the method.The Spark-based stochastic forest algorithm is reduced by 9.9% in the rmse compared with the Spark-based linear regression method in the flow forecast.And because of its distributed characteristics it can deal with large-scale data better compared to the traditional stand-alone algorithm.In addition,the results from VISSIM simulation show that,compared with the traditional fixed timing method,the average delay during the peak period of the intelligent control method proposed in this thesis is reduced by 5.0% on average,and the average queuing length is reduced by an average of 7.5%.The average delay during Non-peak period is decreased by an average of 7.0% and the average queuing length decreased by an average of 6.7%.In conclusion,the intelligent control method of traffic signal proposed in this thesis can complete the whole process of data acquisition,processing analysis and intelligent control of traffic signal.At present,the algorithm has been integrated into the intelligent transportation platform of cooperation partner,and the project has passed the acceptance.