Research On The Traffic States Prediction And Application Of Signalized Intersection

Intersection is the most important node of network, its traffic state is essential. When traffic doesn‘t run smoothly, intersection can easily become a bottleneck. Most urban traffic accidents occur in the intersection. Researching the prediction method and practical application of intersection traffic states can provide decision services for traffic management control and guidance and enrich the theoretical system of traffic state research.In this paper, the "traffic flow forecasting-traffic simulation-status evaluation" traffic state prediction theoretical framework is proposed and applied to the intersection signal control. Firstly, define the meaning of traffic state and clear the status classification criteria. In particular, analyze each state evaluation index, identify similarities and differences, finally establish an effective selection principle and select saturation, delay, queue length as evaluation index. According to the characteristics of intersection traffic flow, select the traffic volume and time occupancy as objects of short-term traffic flow prediction. Study a non-parametric regression model based on k nearest neighbor and compare it with ARMA model. Finally, establish a forecasting model by combining the above two models with particle swarm. Using traffic simulation, multi-attribute decision theory and fuzzy theory build a comprehensive evaluation model of traffic status. Revise the membership function with cluster analysis and merge the subjective weight by preference rate method with objective weight by entropy method in game theory. Study the intersection control strategy based on traffic state and select delay and capacity as optimization objective function of signal timing. Transform the multi-object problem into a single object problem by fuzzy compromise programming method. Construct a weighting expression of function with traffic state comprehensive evaluation model output. Finally solve the model by particle swarm algorithm.The results showed that: the classification criteria of traffic status and the evaluation index can meet the application; the non-parametric regression model based on k nearest neighbor is better than the ARMA model, and the combined model which is the best summing up all the advantages of the models can be used to predict traffic flow of intersection; Using traffic simulation can improve the efficiency of the evaluation, since the full account of subjective and objective information, the membership function and weight vector of optimized evaluation model are more scientific and applicable. Fuzzy compromise programming method can effectively solve the problem of multi-objective optimization signal timing and the function weight calculated by the state comprehensive evaluation model avoids the defect of conventional method which determine the function weights subjectively.Studies have shown that: intersection traffic state prediction theoretical framework proposed in this paper is basically feasible. It can guide the signal control optimization, provide a reference for traffic management, inducing and provide ideas for further study on intersection traffic state.