Traffic Modeling with Multi Agent Bayesian and Causal Networks and Performance Prediction for Changed Setting System

The traffic modeling is one of the effective methods of detecting and evaluating the urban traffic. The effect of uncertain factors such as the different behavior of a human society would count as an intricacy of the issue and would cause some problems for modeling. Level crossroads are one of the important sections in an urban traffic control system and are usually controlled by traffic lights. In this study, an attempt has been made to model the traffic of an important crossroads in Mashhad city using intelligent elements in a multi-agent environment and a large amount of real data. For this purpose, the total traffic behavior at the intersection was first modeled based on the Bayesian networks structures. Then, effective factors have been modeled using the probabilistic causal networks. Results of the evaluation of the model show that this model is able to measure system efficiency according to variances in the crossroads adjustments. Also, this model is cheaper and less time-consuming. On this basis, this modeling can be used for the evaluating and even predicting the efficacy of the traffic control system in the crossroads. The data used in this study have been collected by the SCATS software in Mashhad Traffic Control Center. The Weka software has been used for training and evaluations with the Bayesian and causal probabilistic networks.