Enhancing Road Traffic Flows Across Multiple Roundabouts Using Adaptive Neuro-Fuzzy Inference System

Abstract

One of the classical problems of modern cities is traffic congestions. Their negative effects on health, education, and economy cannot be overemphasized. The Kaduna metropolis, which has eight roundabouts and one underpass between Kakuri and Hayin Banki often experience traffic congestions and efforts, which include un-signalized and static-timed signalization at the roundabouts have proven to ineffective. However, no application of computational intelligence tools, which are dynamic, adaptive and have proven to better at effective scheduling of traffic flows. Therefore, this research proposes and developed traffic controllers based on adaptive neuro-fuzzy inference systems’ clustering methods. The grid partitioned, subtractive clustering and fuzzy C-means clustering controllers were used in a simulation of urban mobility traffic simulation and the results showed that all the ANFIS-based dynamic controllers essentially outperformed the usual 30s and 25s static controller in terms of minimization of vehicular average waiting time. The grid partition controller performed best as it minimized the average waiting time by over 24% over the 30s controller.