Abstract
This study presents EATS-MEC an intelligent MEC framework aimed at optimizing energy-aware task offloading and resource scheduling in ultra-dense network environments. The objective is to improve energy efficiency, scalability, and latency compliance under heterogeneous and mobile edge conditions. EATS-MEC integrates Deep Reinforcement Learning (DRL) for real-time task allocation and a lightweight blockchain module to ensure secure, decentralized execution across edge, fog, and cloud layers. Unlike classical models such as Deep Q-Networks (DQN) and Genetic Algorithms (GA), EATS-MEC adaptively responds to real-time network and mobility feedback to determine the optimal execution location for each task. Simulations demonstrate that EATS-MEC reduces peak energy consumption by 32%, extends device battery life by up to 20 hours, and achieves a task success rate of 88.3% under stringent deadline constraints. The framework shows superior performance in mobility-aware energy usage and exhibits near-sublinear energy scaling behavior with increasing device density, maintaining high task throughput even with over 100,000 concurrent tasks. Results indicate that EATS-MEC outperforms existing baselines in energy-latency trade-offs and operates close to the Pareto frontier. Due to its robust, secure, and adaptive nature, EATS-MEC is highly suitable for deployment in real-world smart city infrastructures, healthcare IoT, and latency-sensitive industrial applications.