Improvement of Adaptive Control of Traffic Flows at a Traffic Light Object

. The article considers methodological aspects related to the adaptive traffic light control system based not on calculating the duration of phases, but on determining the required number of vehicles to be passed in each phase. It is assumed that modern technical capabilities of computer vision and neural networks will be used, which allow applying approaches typical for manual traffic control at intersections. The algorithm is based on constant monitoring and accounting of traffic intensity on the sections.