This research introduces an integrated framework that combines Artificial Intelligence (AI) and Internet of Things (IoT) technologies to enhance public transportation systems. The framework addresses critical urban mobility challenges, such as service reliability, resource utilization, and passenger satisfaction, through a multi-layer architecture designed for intelligent decision-making and optimization. The methodology involves deploying IoT sensor networks across vehicle fleets, transit stations, and traffic intersections to collect real-time data. This data is processed using advanced AI techniques, including deep learning models like Long Short-Term Memory (LSTM) networks for demand prediction and Temporal Convolutional Networks (TCN) for pattern recognition, which improved accuracy by 27.3% compared to traditional methods. Additionally, reinforcement learning algorithms, such as deep q-networks and proximal policy optimization, significantly reduced wait times by 32.8%. The system features a three-tier architecture comprising edge, fog, and cloud layers, ensuring efficient local processing and global optimization with a response latency of under 50 milliseconds and high reliability (99.99%). Results showed substantial improvements: a 24.6% reduction in vehicle idle time, a 42.7% decrease in average waiting time, and a 21.8% reduction in operational costs. This research contributes to the field by introducing adaptive learning algorithms and a scalable deployment architecture. The findings have implications for urban planning, economic efficiency, and social benefits, enhancing public service accessibility and overall user experience in urban transportation systems.