Modeling of vehicle delays at signalized intersection with an adaptive neuro-fuzzy (ANFIS)

Abstract

An adaptive neuro-fuzzy inference based delay estimation system is proposed. The system is compared with other delay estimation models, and tested through simulation and observation values. Rules, fuzzification and inference are modeled by neuro-fuzzy. Hybrid algorithm has been used for training and tests. The rule base of the delay estimation system is constructed either following a mathematical model or from real-time traffic operational data. This study has shown that adaptive neuro-fuzzy technique, a method to predict vehicle delays at signalized junctions, can be successfully applied to modeling of traffic systems.