Research On Online Traffic Flow Distribution Based On Traffic Flow Prediction

It is the focus of the whole society and international academic circles that how to distribute the traffic flow to the limited road resources to make the total travel cost or individual travel cost as little as possible.The existing studies usually assume that user can obtain the determined traffic flow or its distribution in advance,which does not meet the actual demand.Aiming at the problem of real-time traffic flow assignment of the starting node to the destination node in the future uncertain traffic flow,this paper adopts the theory and method of online problem and competitive strategy on the basis of the prediction of the uncertain traffic flow in the future.Also,it can provide effective theoretical basis for the real-time routing selection that the traffic management departments direct traffic and user’s travel in terms of online traffic flow distribution strategy under the condition of unlimited and limited road capacity from the perspective of traffic managers and users.The main contents and innovative works in this paper Are as follows.Design and competitive performance analysis of online user equilibrium（UE）strategy based on Bayesian traffic flow prediction.The travel time and travel volume of each individual user is uncertain from the point of view of individual users.Therefore,the traffic flow of a certain starting node in a certain period of time is uncertain.The paper first finds out the user’s current traffic flow to reach the starting node using the Probit method.On this basis,according to the characteristics that the individual users can only obtain the limited traffic flow data,the neural network model,the Kalman filter model and the nonparametric regression model are used to form the Bayesian combination forecasting model A,which forecasts the traffic flow for the next period of time.Then we adopt the theory and method of online problem and competitive strategy in this paper,and design the Bayesian online user equilibrium strategy with the aim of each user online travel cost as little as possible which distributes traffic flow to the road network based on Bayesian flow prediction.It is proved that the proposed strategy can be used to analyze the competitive ratio and effectiveness of two kinds of situations.Also,the competitive ratio is（1+（α+β）/2）^g+2 under the condition that the road capacity is unlimited,and the competitive ratio under the limited road capacity is（1+λ（α+β）/2）^g+2.An example is analyzed to verify the implementation effect of Bayesian online UE strategy.Design and competitive performance analysis of online system optimization（SO）strategy based on Bayesian traffic flow prediction.According to the characteristics of the traffic flow data of each time period and each section from the traffic managers’ point of view,time series-ARIMA model,support vector machine regression model and historical average model are used to form the Bayesian combination forecasting model B,which forecasts the traffic flow for the next period of time.Then we adopt the theory and method of online problem and competitive strategy in this paper,and design the Bayesian online system optimum strategy with the aim of all users’ online travel cost as little as possible which distributes traffic flow to the road network based on Bayesian flow prediction.It is proved that the proposed strategy can be used to analyze the competitive ratio and effectiveness of two kinds of situations.Also,the competitive ratio is（1+（α+β/2））^g+1 under the condition that the road capacity is unlimited,and the competitive ratio under the limited road capacity is（1+λ（α+β）/2）^g+1.An example is finally analyzed to verify the implementation effect of Bayesian online SO strategy.