Intelligent information system for urban traffic flow management based on a multi-level agent architecture

Abstract

This paper proposes the concept of an intelligent information system for urban traffic flow management based on a multi-level agent architecture and artificial intelligence technologies. The relevance of the study is driven by the need to improve the efficiency of transportation data processing, provide real-time decision support, and ensure adaptive traffic flow management under conditions of increasing pressure on urban infrastructure. The proposed information system implements a multi-level agent interaction model that includes local, regional, and global man agement levels, providing the collection, integration, analysis, and processing of transportation environment data with the subsequent generation of control decisions. The paper examines the functional structure of the system, the principles of inter-level agent coordination, traffic flow forecasting mechanisms, and decision-support algorithms. To predict the dynamics of transportation processes, artificial intelligence models, including LSTM, GRU, and Transformer neural networks, are employed, enabling the consideration of temporal dependencies and fluctuations in traffic intensity. To evaluate the effectiveness of the proposed information system, a series of simulation experi ments was conducted using the SUMO, AnyLogic, and Aimsun environments based on the transport network of the city of Lviv. Experimental studies were carried out for three typical scenarios of urban transport network oper ation: peak traffic periods, emergency situations, and route congestion. The obtained results demonstrated that the use of the proposed multi-level agent-based information system reduces average vehicle delay, increases average travel speed, decreases carbon dioxide emissions, and lowers energy consumption compared with conventional traffic management approaches. A comparative analysis with modern systems, including SURTRAC, MADRL, and KS-DDPG, confirmed the advantages of the proposed approach in terms of scalability, adaptability, multi criteria optimization, and the possibility of integration into the Smart City ecosystem. Future research directions are related to the further development of decision-support mechanisms, integration of V2X technologies, and expansion of the information system functionality for autonomous transportation environments.