A vehicular traffic congestion predictor system using Mamdani fuzzy inference

Abstract

The process of traffic control systems significantly relies on the immediate detection of breakdown states. As a result of their crisp (non-fuzzy) based calculation procedures, conventional traffic estimators and predictors cannot effectively model traffic states. In fact, these methods are characterized by exact features, while traffic is defined by uncertain variables with vague properties. Furthermore, typical numerical methodologies have constraints on evaluating the overall system status in heterogeneous and convoluted networks mainly due to the absence of reliable and real-time data. This study develops a fuzzy inference system that uses data from the Hungarian freeway networks for predicting the severity of congestion in this complex network. Congestion severity is considered the output variable, and traffic flow along with the length and the number of lanes of each section are assigned as input variables. Seventy-five fuzzy production rules were generated using accessible datasets, percentile distribution, and experts' consensus. The MATLAB fuzzy logic toolbox simulates the designed model and analysis steps. According to available resources, the results demonstrate linkages among input variables. Analyses are also used to construct intelligent traffic modeling systems and further service-related planning.