Balancing Energy and Latency in Multi-Edge MEC Systems Using DQN-Based Task Offloading

The rapid growth of IoT devices and mobile applications has increased the need for efficient Mobile Edge Computing (MEC) systems that optimize task offloading and resource allocation. This paper presents a Deep Q-Network (DQN)-based approach for managing these processes in a multi-edge MEC environment with multiple users. Using deep reinforcement learning (DRL), we evaluate three reward functions that prioritize energy efficiency, task completion time, or a balance between both. Experimental results demonstrate the trade-offs between these objectives. The energy-focused function minimizes consumption, making it suitable for energy-constrained environments such as UAV networks. The time-focused function achieves faster execution, benefiting latency-sensitive applications like real-time monitoring. The balanced function provides a middle ground, optimizing both energy efficiency and execution time. These insights contribute to the design of adaptable MEC systems, particularly in smart city applications such as traffic management and environmental monitoring.