Adaptive neuro-fuzzy traffic signal control for multiple junctions

Abstract

The performance of neuro-fuzzy traffic signal control at two independent traffic junctions is discussed. In this work, traffic conditions at two 4-way traffic junctions were simulated and flow of traffic on the road connecting the two junctions under varying traffic conditions was studied. For a given data set, the developed neuro-fuzzy system automatically draws membership functions and the rules by itself, thus making the designing process easier and reliable compared to conventional fuzzy logic controllers. The traffic inflows of roads are given as inputs to the fuzzy control system which generate the corresponding green light time as the output to control the signal timing. The control systems try to minimize the delay experienced by the drivers at the two traffic junctions. As expected, when compared with a fixed-time signal control system, the neuro-fuzzy system tends to minimize vehicle delays at both junctions. Simulation results show, under variable traffic conditions, neuro-fuzzy control system perform efficiently by making average delay per vehicle under the red light phase smaller and increasing the synchronization of green light phases between junctions. When the volume of traffic at one of the junction changed abruptly, the green light timing of both junctions changed, adapting to the new traffic condition on the road connecting the two junctions.