Real-Time IoT-Based Smart Traffic Management System Using Edge AI and Sensor Networks

Abstract

The fast increase in vehicles in cities has brought about serious problems with crowded roads, extended trips and more pollution, prompting the invention of intelligent traffic control systems. The research proposes a smart traffic management system built using IoT infrastructure, edge AI and a spread of sensors that aim to optimize traffic flow, decrease delay at traffic signals and help the environment. Split computing framework makes it possible for decisions related to traffic lights to be made locally, not through the cloud. Data from a mixture of inductive loop detectors, RFID vehicle identifiers, sensors for air quality and IP cameras is sent in real time. For edge deployment, a light CNN is trained to spot vehicles and estimate how many cars are on each lane, while LSTMs predict minor changes in traffic flow to allow managers to plan ahead. Otherwise, agents with reinforcement learning respond using Deep Q-Learning which enables the system to alter traffic signal phases automatically when traffic or conditions change. The system was tested by running simulations and deploying hardware with Raspberry Pi which led to clear increases in how well the system runs. Average vehicle idle time was reduced by 27.2%, intersection throughput increased by 21.1% and CO₂ emissions decreased by 18.1% when these adaptive signals were used in place of conventional ones. The results show that the proposed method can handle large volumes, respond in real time and be maintained which makes it a good fit for future traffic management systems in smart cities.