Scholarlink

<p>In future sixth-generation (6G) communication networks, the Internet of Things (IoT) is expected to play a crucial role, necessitating energy-efficient computation despite challenges such as limited battery life and resource constraints. Mobile edge computing (MEC) offers a promising solution by providing on-demand computation capabilities to low-power IoT devices. However, as the number of IoT devices increases, congestion and performance degradation can occur in single-MEC server systems. This study proposes a novel multi-MEC network architecture that leverages a realistic path loss model and a dynamic offloading strategy, incorporating both partial and binary offloading schemes to enhance computational energy efficiency. We formulate a joint optimization problem that addresses user association, offloading volume, offloading strategy, and the allocation of computational and communication resources. To solve this complex problem, we employ a multi-agent deep reinforcement learning (MADRL) approach using the multi-agent deep deterministic policy gradient (MADDPG) algorithm. Extensive simulations demonstrate that our proposed approach outperforms traditional algorithms such as deep Q-network (DQN), proximal policy optimization (PPO), and asynchronous advantage actor-critic (A3C), as well as conventional binary and partial offloading strategies. Numerical results reveal that our scheme improves performance by approximately 22.01% compared to the binary scheme and 8.26% compared to the partial offloading scheme. Furthermore, statistical analysis shows the effectiveness of the proposed framework, reducing outage probability to 0.01% at 30 dBm and achieving a task completion ratio of 99% with the deployment of 18 MEC servers. These results confirm the potential of MADRL-based methods to optimize resource allocation and enhance computational efficiency, paving the way for more resilient and effective frameworks to address the constraints of low-power IoT devices in future communication networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing computational efficiency in 6G IoT networks: a multi-agent deep reinforcement learning approach for multi-MEC systems

  • 期刊论文

  • 发表日期：

    2025年11月22日

查看全文
  • Abdulbasit A. Darem,
  • Asma A. Alhashmi,
  • Fawaz Alanazi,
  • Tareq M. Alkhaldi,
  • Abed Saif Ahmed Alghawli,
  • Khaled M. Alalayah,
  • Monir Abdullah

摘要

In future sixth-generation (6G) communication networks, the Internet of Things (IoT) is expected to play a crucial role, necessitating energy-efficient computation despite challenges such as limited battery life and resource constraints. Mobile edge computing (MEC) offers a promising solution by providing on-demand computation capabilities to low-power IoT devices. However, as the number of IoT devices increases, congestion and performance degradation can occur in single-MEC server systems. This study proposes a novel multi-MEC network architecture that leverages a realistic path loss model and a dynamic offloading strategy, incorporating both partial and binary offloading schemes to enhance computational energy efficiency. We formulate a joint optimization problem that addresses user association, offloading volume, offloading strategy, and the allocation of computational and communication resources. To solve this complex problem, we employ a multi-agent deep reinforcement learning (MADRL) approach using the multi-agent deep deterministic policy gradient (MADDPG) algorithm. Extensive simulations demonstrate that our proposed approach outperforms traditional algorithms such as deep Q-network (DQN), proximal policy optimization (PPO), and asynchronous advantage actor-critic (A3C), as well as conventional binary and partial offloading strategies. Numerical results reveal that our scheme improves performance by approximately 22.01% compared to the binary scheme and 8.26% compared to the partial offloading scheme. Furthermore, statistical analysis shows the effectiveness of the proposed framework, reducing outage probability to 0.01% at 30 dBm and achieving a task completion ratio of 99% with the deployment of 18 MEC servers. These results confirm the potential of MADRL-based methods to optimize resource allocation and enhance computational efficiency, paving the way for more resilient and effective frameworks to address the constraints of low-power IoT devices in future communication networks.