Collaborative MEC\u2013Metro Edge for Mobility-Aware Task Offloading in Smart Cities Using Reinforcement Learning

The growing number of mobile devices in smart cities generates massive data streams that require low-latency and energy-efficient processing. \\ac{mec} brings computation closer to users; however, single-tier \\ac{mec} often fails under high mobility and workload. This limitation results in congestion, increased costs, and task failure. To address this gap, we propose a mobility-aware collaborative \\ac{mec}-Metro edge offloading framework using \\ac{rl}. To evaluate our study, we developed \\textit{SmartCityEdgeSimPy}, a purpose-built simulator that models a smart city of 10$\\times$10 km with realistic mobility and diverse workloads. The simulator enables visualization of offloading decisions and records detailed performance metrics. Results show that the framework sustains \\ac{mec} utilization between 78-90\\%, activates \\ac{mec} collaboration for about 15\\% tasks, and significantly reduces delay, cost, and energy compared to Metro and Cloud. These outcomes confirm that the proposed framework ensures scalability and efficiency in mobility-aware smart city environments.