Reinforcement Learning-Driven Task Offloading: Improving MEC Efficiency with DQN and DDPG

Abstract

The rapid evolution of edge computing, particularly Mobile Edge Computing (MEC), has prompted the need for efficient task offloading strategies to optimize network resources, energy consumption, and latency. As applications requiring low latency and high data processing capabilities proliferate, offloading computational tasks from mobile devices to edge servers has become essential. This paper explores the use of Deep Reinforcement Learning (RL) models, specifically Deep Q-Networks (DQN) and Deep Deterministic Policy Gradient (DDPG), to optimize task offloading in MEC systems. We analyze a multi-user, single-server scenario and compare the performance of DQN and DDPG in reducing energy consumption and delay. Our results demonstrate that DQN outperforms DDPG in terms of reward stability, energy efficiency, and latency management, making it more suitable for real-time applications. The study highlights the potential of RL strategies to improve MEC performance and suggests future research on multi-server environments.