APPLICATION OF NEURAL NETWORK TECHNOLOGIES FOR CONTROL OF TRAFFIC LIGHTS

The paper describes the use of an artificial neural network to determine the optimal parameters of traffic light regulation based on the intensity of traffic flow. At regulated intersections, there is an imbalance in the intensity of traffic flow, due to which one operation mode of traffic lights at an intersection may be ineffective. The study objective is to develop software for predicting the operating modes of traffic lights, taking into account the spatial and temporal unevenness of transport demand. Based on the simulation of traffic flows at one regulated intersection, the values of the average delay time were determined for different traffic light operating modes and traffic flow intensities, including turning ones. The artificial neural network was trained on data from 16 thousand simulations and tested on four thousand simulations. Using an artificial neural network to calculate the optimal operating mode of traffic lights reduces the delay time by 20-50% for two rush hours. A pre-trained artificial neural network can calculate the optimal operating mode of traffic lights for a specific regulated intersection in one second. The developed software can be used to implement an intelligent transport system in an automated traffic control system.