Abstract
A modern system of intelligent dispatching for a complex road infrastructure facility in the metropolis of Almaty is being presented. To improve the existing automated traffic management system (ATMS) in the city and to gradually implement digital traffic technologies with autopilot vehicles, a crucial and pressing task for IT is to replace the traffic police dispatcher-employee with a neural network-based artificial intelligence. The work demonstrates the methods and results of implementing a neural network into a mobile server program for the AnyLogic environment using the SUMO application. This approach, using the built-in TraCI connector, established connections with Python scripts of the server program AnylogicRaspberryServer.py. Testing and analysis of the research results showed the adequacy of virtual dispatching, which allowed reducing the load on the traffic police employee under moderate and average traffic conditions. It is noted that manual control of the situation at the Abay/Saina intersection is more preferable during peak hours. The obtained software products and the semi-industrial prototype of the local ISU with AI are protected by copyright certificates and have important practical significance.