Digital twin of an intersection for adaptive traffic light phase control

The paper presents the development of an adaptive traffic control system (ATCS) for a single intersection, aimed at increasing its throughput. The proposed solution is a digital prototype of an intelligent ATCS implemented in Python and equipped with a graphical user interface (GUI) for monitoring and adjustment. A computer vision module based on the YOLOv4-Tiny neural network was integrated for collecting data on vehicles and pedestrians. The system implements a control algorithm that adjusts traffic light phase timings in real time according to three designed scenarios: night mode, peak hours, and medium load. For comparative efficiency analysis, a digital twin (DT) simulation model was developed in AnyLogic. The simulation included throughput assessment under different scenarios. The results showed that the introduction of adaptive control algorithms reduces the average intersection travel time by 14–32% compared to traditional static algorithms.