DEVELOPMENT OF AN AI-DRIVEN CONTROL ARCHITECTURE FOR REAL-TIME TRAFFIC OPTIMIZATION AND CYBER THREAT DETECTION IN SMART CITIES.

This paper presents the development of an AI-driven control architecture designed for real-time traffic optimization and cyber threat detection in smart city Cyber-Physical Systems (CPS).The proposed framework integrates Support Vector Regression (SVR) for traffic efficiency prediction and Support Vector Machine (SVM) for cybersecurity anomaly classification.Simulation results using a synthetic urban traffic dataset revealed that SVR achieved a low Root Mean Squared Error (RMSE) of 0.2861, demonstrating high prediction accuracy for traffic reward outputs, while SVM achieved 90% classification accuracy with a 61% detection rate, indicating robust cyber threat detection capabilities.Comparative analysis against traditional fixed-timing traffic control and heuristic-based intrusion detection systems confirmed the superior adaptability and security performance of the proposed AI models.Engineering implications highlight improvements in traffic flow management, operational resilience, and urban safety.Limitations include reliance on synthetic datasets and computational demands of real-time deployment.Future work will focus on integrating explainable AI modules, hybrid reinforcement learning architectures, and hardware-in-the-loop validation for deployment in large-scale smart city infrastructures.