An artificial intelligent system for traffic forecasting in virtual stations of highways

Abstract

Traffic data and statistics (as density) are among the most important and crucial tools for the study of traffic, optimization of urban transportation and traffic programming. For this reason, the Tehran Traffic Control Company (TTCC) has built some data acquisition stations in various highway locations in Tehran. There are TTCC 10 data acquisition stations in use at the moment and more stations are to be designed and implemented in the future. The implementation of each station consists of three stages: field study, equipment installation and setting up magnetic rings, which require a long design time and high cost for data acquisition equipment, as well as induction sensors and heavy work load for installing magnetic rings. The final stage comes with many problems, particularly in highways, and needs high safety standards. Considering problems such as cost and safety of the implementation, the design of virtual stations seem necessary. A virtual station does not exist physically, but it can calculate traffic data. This proposal is known as Spatial Forecasting, which provides forecasting in space dimension using statistical data from present stations. The present article introduces an intelligent neuro-fuzzy structure which is trained over statistical traffic density data of the Modarres highway and is used for spatial forecasting. Results of simulation show that a neuro-fuzzy network can be used successfully for this purpose