Modeling vehicle delays at signalized junctions : Artificial neural networks approach

Delay of vehicles at signalized junctions is one of the main criteria used for evaluation of control systems' performances. The vehicle delay is uniform and non-uniform delay types. The uniform part consists of signal timings; the non-uniform part includes vehicle queuing, random arrivals and over-saturation cases of traffic flows. The uniform part of the vehicle delays is basically determined using conventional delay formulas. But for the non-uniform part, artificial neural network (ANN) approach is used and a vehicle delay estimation model [artificial neural network delay estimation of traffic flows (ANNDEsT)] is developed. ANNDEsT model compared with Webster, HCM and Akcelik delay calculation methods and field observations, shows encouraging results especially for the cases of over-saturation or non-uniform conditions.