A new approach to model traffic light controller (TLC) relying exclusively on Artificial Neural Networks (ANN) technology; we present as well a comparative approach of ANN architectures applied to solve the underlying traffic problem. Traditional controllers have limitations and cannot adapt to changing traffic demands, where adaptive controlling requires mathematical modelling and optimisation where traditional methods is insufficient for modelling and controlling the system due to non-linearity and nondeterministic nature of traffic control model. Artificial intelligence techniques were highly utilised to deal with problems of similar category, and proves superiority in TLC applications. In this work, we explore the generalisation capability of various ANN models in solving the TLC problem. The performances of different models are compared by analysing the training and testing results of each network.