At present urban traffic congestion continues to grow as cities expand, and Intelligent Transportation Systems (ITS) increasingly rely on Internet of Things (IoT) sensors and edge devices to monitor and manage traffic in real time. Two capabilities are central to such systems: adaptive control of traffic signals to reduce congestion, and rapid detection of traffic anomalies such as accidents or sudden congestion events. Existing research typically treats these two capabilities in isolation and further assumes unconstrained cloud-level computing resources, which does not reflect the reality of edge-deployed IoT hardware with limited compute, memory, and energy budgets. This project proposes a resource-constrained framework that jointly integrates
(i) Reinforcement-learning-based adaptive traffic signal controller and
(ii) Anomaly-aware edge-cloud offloading strategy for real-time traffic anomaly detection, evaluated at the level of a small, realistic intersection network.
The framework will be developed and evaluated using the SUMO traffic simulator with a real-world-calibrated scenario, combined with real sensor datasets and incident records for anomaly detection. The project targets a measurable improvement in traffic responsiveness and anomaly-handling speed over static baselines, while remaining within realistic edge-device resource budgets, and contributes an integrated optimization framework that has not been jointly addressed in prior work.