Neuro-Fuzzy Traffic Signal Control in Urban Traffic Junction

This work is based on the implementation of new concepts for adaptive traffic signal controller using a neuro-fuzzy system approach and simulations on reference test cases. In our neuro-fuzzy controller, the parameters of the fuzzy membership functions are adjusted using a neural network. The neural learning algorithm may then be considered as reinforcement learning. However, the major difficulty for this neuro-fuzzy system under consideration is such that the most usual neural learning algorithms cannot be used. A specific learning algorithm is proposed to be used both for constant traffic volumes and also for changing volumes. Starting from the initial membership functions, the learning algorithm modifies the parameters of the membership functions in different ways at different but constant traffic volumes. The membership functions after the proposed learning algorithm produce smaller delays than the initial membership functions. An additional contribution is for specific changes in the rule base of the fuzzy traffic signal controller in order to reduce delays in various traffic volumes conditions in a test/reference traffic junction.