Published Online:October 2026
Product Name:The IUP Journal of Computer Sciences
Product Type:Article
Product Code:IJCS031026
DOI:10.71329/IUPJCS/2026.20.4.43-59
Author Name:Manush Dip Budhathoki, Rahul Kumar Yadav, Niharika Srivastava, Sagar K G and Sunena Rose M V
Availability:YES
Subject/Domain:Engineering
Download Format:PDF
Pages:43-59
The paper presents an in-depth synthesis and critical evaluation of simulation-based and empirical methods for artificial intelligence (Al)-enabled adaptive traffic control systems (ATCS) in various urban settings. The analysis draws on peer-reviewed studies published from 2020 to 2025. It covers meta-analytical investigations, cloud-based deep learning frameworks, multi-agent system architectures, large language model (LLM)-integrated simulation tools, graph neural network (GNN) forecasting models, federated learning (FL) deployments, and Internet of Things (IoT)- driven smart city implementations. The simulation platforms evaluated include VISSIM, SUMO, MATSim, AIMSUN, and LLM-assisted ChatSUMO framework. The AI techniques considered consist of reinforcement learning (RL), deep Q-networks (DQN), multi-agent RL (MARL), long short-term memory (LSTM), gated recurrent units (GRU), and graph-based deep learning models. The key performance indicators, such as travel time index (TTI), vehicle delay, queue length, and predictive error metrics (MAE, RMSE, MAPE), are summarized to allow comparisons across different methods. New approaches, including GNNs, FL, connected and autonomous vehicles (CAVs), and privacy-preserving designs, further enhance the capabilities of smart traffic systems. The paper concludes by outlining a Web-based deployment pathway to make simulationdriven traffic optimization tools accessible to everyone.
Urban traffic congestion imposes severe economic, environmental, and social costs on cities across the globe.