A.D. Sovastyuk1, A.V. Katkov2

1-2 OJSC "AGAT-Control Systems" – Managing Company of "Geoinformation Control Systems" Holding (Minsk, Belarus)

1sovastyuk-ad@agat.by, 2katkov@agat.by

This article addresses the pressing issue of improving the efficiency of traffic management in the context of rapid growth in the number of vehicles and a significant lag in the development of transport infrastructure. It examines the evolution of traffic management systems from traditional statistical methods to promising intelligent solutions based on artificial intelligence (AI). The aim of this work is to analyse existing intelligent transport systems (ITS) and to justify the prospects for applying neural network methods to the prediction and scenario modelling of traffic flows.

The first part of the work presents an overview of three generations of ITS. The first generation (eCall, TMC) is characterised by reactive information provision without the use of machine learning. The second generation (adaptive SCOOT/SCATS systems and the Automated Traffic Control System) provides dynamic traffic light control, but uses heuristic algorithms with limited integration of AI. The third generation (digital platforms such as Google Maps and Yandex Maps) implements congestion forecasting based on user data; however, it is not directly integrated with the road infrastructure and does not allow for the modelling of the consequences of management decisions.

The main part of this work is devoted to the mathematical formulation of two key problems. Firstly, prediction traffic conditions up to 24 hours in advance, where the use of recurrent neural networks (LSTM and GRU) is proposed as the basic tool to account for long-term temporal dependencies. To overcome the limitations of classical RNNs (which do not take into account the spatial structure of the network), the use of graph neural networks (GNNs) is proposed, which represent the road network as a graph. Secondly, scenario-based simulation modelling (what-if analysis) is proposed to assess the consequences of road traffic accidents, road closures or changes to traffic patterns, for which the open-source SUMO platform is recommended. In conclusion, the implementation of the proposed approach will improve the soundness of management decisions and reduce congestion and accident rates.

Sovastyuk A.D., Katkov A.V. The use of artificial intelligence in traffic management // Information-measuring and Control Systems. 2026. V. 24. № 4. P. 50−56. DOI: https://doi.org/10.18127/j20700814-202604-06

1. Meyer M.D., Miller E.J. Urban transportation planning: a decision-oriented approach. New York: McGraw-Hill. 2001.
2. Ortúzar J.D., Willumsen L.G. Modelling transport. Chichester: Wiley. 2011.
3. McNally M.G. The four-step model. Handbook of transport modelling. Oxford: Elsevier. 2000. P. 35−53.
4. Box G.E.P., Jenkins G.M., Reinsel G.C. Time series analysis: forecasting and control. Hoboken: Wiley. 2015.
5. Ahmed M.S., Cook A.R. Analysis of Freeway Traffic Time-Series Data by Using Box-Jenkins Techniques. Transportation Research Record. 1979. № 722. P. 1−9.
6. World urbanization prospects. New York: United Nations. 2018.
7. Baranov A.V., Gusev S.A. Intellektualnye transportnye sistemy: analiz i perspektivy razvitiya. Informatsionnye tekhnologii i vychislitelnye sistemy. 2019. № 3. S. 45−58. (in Russian)
8. Afanasev A.A., Lebedev I.S. Primenenie metodov iskusstvennogo intellekta v transportnykh sistemakh. Transport Rossiiskoi Federatsii. 2021. № 2. S. 33−41. (in Russian)
9. Kovalev V.V., Sinitsyn I.N. Modelirovanie transportnykh potokov v gorodskikh setyakh. Vestnik MADI. 2020. № 4. S. 15−27. (in Russian)
10. Intelligent transport systems. eSafety. Pan-European eCall operating requirements: BS EN 16072:2025. London: British Standards Institution. 2025. 34 p.
11. 511: America’s Traveler Information Telephone Number. Washington. D.C.: U.S. Department of Transportation. https://www.fhwa.dot.gov/trafficinfo/511what.htm (15.06.2026).
12. TMC Location Codes (road traffic). Open Transport Data Switzerland. https://opentransportdata.swiss/en/cookbook/road-traffic-cookbook/tmc-location-codes-strassenverkehr/ (15.06.2026).
13. Hunt P.B., Robertson D.I., Bretherton R.D., Winton R.I. SCOOT – A Traffic Responsive Method of Coordinating Signals. Crowthorne: Transport and Road Research Laboratory. 1981. 198 p.
14. Lowrie P.R. SCATS: Sydney Coordinated Adaptive Traffic System – A traffic responsive method of controlling urban traffic. Sydney: Roads and Traffic Authority of New South Wales. 1982. 160 p.
15. Gosudarstvennoe kazennoe uchrezhdenie goroda Moskvy "Tsentr organizatsii dorozhnogo dvizheniya". Intellektualnaya transportnaya sistema goroda Moskvy. M.: TsODD. https://transport.mos.ru/ (15.06.2026). (in Russian)
16. Hochreiter S., Schmidhuber J. Long Short-Term Memory. Neural Computation. 1997. V. 9. № 8. P. 1735−1780.
17. Li Y., Yu R., Shahabi C., Liu Y. Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting. International Conference on Learning Representations (ICLR). 2018.
18. Roosta A., Kaths H., Barthauer M., Erdmann J., Behrisch M. Microscopic traffic simulation using SUMO. IEEE Intelligent Transportation Systems Conference (ITSC). 2018.