Adaptive traffic signal control is an important requirement for reducing urban congestion and improving traffic flow in smart cities. Traditional fixed-time signal systems work on pre-defined schedules and cannot respond effectively to sudden changes in traffic demand, peak-hour congestion, road incidents, or uneven lane usage. This research paper presents an intelligent traffic signal control system that combines Reinforcement Learning (RL) with vehicle-mounted Internet of Things (IoT) sensors. In the proposed system, vehicles provide anonymized and aggregated traffic information such as position, speed, lane approach, queue formation, and movement direction. This information is collected by roadside aggregation units and used by reinforcement learning agents to dynamically select signal phases at intersections. The main objective of the system is to reduce average waiting time, queue length, unnecessary stops, vehicle idling, and unfair lane delays while maintaining data privacy. A multi-agent Advantage Actor-Critic based approach is considered for controlling multiple intersections, and other RL algorithms such as Q-learning, Deep Q-Network, and Proximal Policy Optimization may also be applied depending on the traffic environment. The system is evaluated through SUMO-based traffic simulation. The study shows that RL-based signal control can improve performance compared with fixed-time and threshold-based control methods, with preliminary simulation results indicating approximately 30% improvement in waiting time and queue length. The paper also discusses methodology, deployment process, scalability, communication challenges, privacy protection, limitations, and future scope of RL-IoT based intelligent traffic management.