Intelligent Control of Urban Intersection Traffic Light Based on Reinforcement Learning Algorithm

The increasing number of vehicles, followed by traffic congestion, has posed a great challenge to the optimal control of traffic for human societies. Therefore, in order to achieve sustainable development in the field of integrated urban management, control of transportation networks is inevitable. The proper method for optimal traffic control should certainly be adaptable in order to be able to manage urban traffic that has a dynamic, complex and changeable nature. In this regard, the method of reinforcement learning that does not require a mathematical model of the environment is very important. In this paper, an intelligent method for controlling urban traffic based on reinforcement learning is presented in which a 4-way intersection is modeled with two different scenarios for low and high traffic congestion. The results obtained after repeated experiments of implementing the proposed method and also its improved model on the mentioned intersection show that the amount of travel time delay has been reduced compared to the usual fixed time methods. After comparing with the two fixed time methods, the waiting time of vehicles at the intersection is 15% and 86% improved for the scenario with low and high traffic congestion respectively, compared to the first method and 37% and 16% compared to the second method.