Applying Adaptive Network-Based Fuzzy Inference System to Predict Travel Time in Highways for Intelligent Transportation Systems

Travel time is a good criterion in analyzing transportation systems. Advanced systems of collecting traffic data (e.g. loop detectors, video cameras …) are now collecting and storing daily status of traffic throughout the world. There are two ways to calculate travel time: direct measurement, and prediction. Several classic statistical ways have been used to predict travel time, but when nonlinear nature is focused, developing a proper model with multiple linear will be a failure. This means that when data have a nonlinear inherent, using of linear methods such as some statistics methods will not be benefit and will not generate appropriate results. Meanwhile, ANN and ANFIS are nonlinear tools. Intelligent systems approaches such as artificial neural networks (ANN) and recently neuro-fuzzy have successfully appeared in prediction. In most applications of ANN, multilayer perceptron (MLP) is applied which is trained by the algorithm of back propagation error. The main problem of this approach is that it is hard to interpret the knowledge in the trained networks. Applying neuro-fuzzy approach, information saved in trained networks will be defined within a fuzzy data base. The aim of present research is to offer a strong neuro-fuzzy network and apply it to predict travel time and compare its results with methods like ANN and AIMSUN. Our results indicate that means for neuro-fuzzy prediction remarkably decrease the error criteria of predicted travel time. This research proves the possibility of applying adaptive neuro-fuzzy inference system in predicting travel time, and reveals that it can make very successful analysis on traffic data. To study credibility of prediction results, AIMSUN traffic simulation software as an expert analyst, was applied and freeway travel time was studied and calculated by simulation.