A Distributed Approach for Coordination Between Traffic Lights Based on Game Theory

Traffic signal control agent can improve its control ability by using the NNQ-learning method. This paper proposes a neural network Q-learning approach with fuzzy reward designed for online learning of traffic lights behaviors. The Q-function table usually becomes too large for the required state/action resolution. In these cases, tabular Q-learning needs a very long learning time and memory requirements which makes the implementation of the algorithm impractical, in real-time control architecture. We considered the problem of coordinating three traffic signals control. The coordination is done using control agents and the concept of game theory. To test the efficiency of the coordination mechanism, a prototype traffic simulator was programmed in visual Studion.net. Results using cooperative traffic agents are compared to results of control simulations where non-cooperative agents were deployed. It indicated that the new coordination method proposed in this paper is effective