Neuro-Fuzzy Green-Time Allocation for Oversaturated Four-Way Signalized Intersections

Abstract

Conventional fixed-time traffic signals allocate a constant green duration regardless of the fluctuation of vehicle volume across approaches, and this rigidity is a common source of congestion at signalized intersections. This study designs and simulates an adaptive traffic light controller based on an Adaptive Neuro-Fuzzy Inference System (ANFIS) with two inputs, queue length (Q) and waiting time (W), to predict the optimal green duration at a four-way intersection. The training data are not arbitrary; they are generated from a queue-actuated control formula grounded in traffic engineering, combining queue discharge time, startup lost time, and an anti-starvation priority term based on normalized waiting time. The model, using three Gaussian membership functions per input, nine first-order Sugeno rules, and hybrid learning, was trained and validated on an 80:20 data split and reached a testing RMSE of 2.39 s and an MAE of 1.87 s. Performance was then evaluated through a 100-cycle discrete-event simulation with Poisson vehicle arrivals and compared against fixed-time control at a constant 30 s green, using an identical queue-update mechanism to keep the comparison fair. Across all seven arrival scenarios the proposed controller produced a lower average total queue, with improvements ranging from 0.8% under light traffic to 87.1% under heavy symmetric traffic. The largest gains occurred near and beyond saturation, where fixed-time control exhibited unbounded queue growth while the proposed controller kept the queue stable through adaptive green-time reallocation.