Optimization of adaptive traffic light control modes based on machine vision

Abstract

Urbanization leads to a significant increase in traffic density in large cities. The growing transport concentration is accompanied by an increase in traffic congestion and emissions of harmful substances. The policies and decisions developed by the authorities, such as the expansion of existing roads and construction of new roads, as well as increase in transport taxes, no longer make it possible to maintain the adequate mobility of the population. One of the solutions is to increase the efficiency of road infrastructure utilization by forecasting the traffic situation and using the adaptive adjustment of traffic lights operation. With dynamic collection and interpretation of traffic flow data from traffic monitoring cameras, it will be possible to use the dynamic parameters of vehicles as indicators for the adaptive adjustment of traffic light regulation. The first part of the paper describes the use of machine vision and a neural network (YOLOv4) in tracking the parameters of traffic flows on road sections in front of intersections. The second part of the paper presents a methodology based on the dynamic regulation of traffic lights cycles and their duration, taking into account the current and forecast parameters of the traffic flow. The algorithm for the adaptive adjustment of traffic lights regulation considers the following parameters: the number and dynamic dimensions of vehicles moving towards the intersection; the number of vehicles in the queue in front of the stop line and their acceleration at the start of the movement. We determined the relationship of the intersection capacity when driving straight ahead with the dynamic dimensions of vehicles and the formation of a queue in front of the stop line, waiting for the green light. The study resulted in the development of an algorithm for setting the duration of both a particular phase and the entire traffic lights cycle in the tasks of eliminating or minimizing the possibility of congestion.