Biologically-inspired neural network for traffic signal control

Urban mobility is a central concern of large cities around the world. The growing urbanization indicates the situation can be even worse. A traffic demand higher than the urban capacity generates traffic congestions, which can be reduced through an efficient traffic signal control method. This paper presents a Biologically-Inspired Neural Network for traffic signal control. Instead of focusing on the macroscopic optimization of urban traffic, like other works, the hereby proposed control method investigates a single intersection between streets. This way, it is possible to incorporate more knowledge about the system dynamics into the control model and analyze its effects on control efficiency. The proposed method consists of a competitive neural network, which balances feedforward and feedback inhibition to synchronize the activity of the neurons, and, thus, the semaphore activation. Moreover, other proprieties of biological neurons are adopted: intrinsic plasticity, to impose system constraints; and synaptic plasticity, to prioritize traffic flows. The flexibility of the neurons and its synaptic connections regarding parameter definition constitute the capacity of easily incorporating knowledge about the system dynamics into the control model. Results of comparative simulations validate the proposed method and illustrate its efficiency and consistency.