Scholarlink

<p>Cluster-based routing protocols play a crucial role in enhancing the applicability of wireless sensor networks (WSNs), which have been facing the challenge of maximizing network lifetime. Addressing this problem, this paper proposes a distributed, energy efficient and balanced clustering and routing protocol called DFDRL which integrates fuzzy logic with deep reinforcement learning (DRL). DFDRL selects optimal cluster heads (CHs) using fuzzy logic based on residual energy ratio, distance deviation, and node degree variance, which achieves intra-cluster energy efficiency and balance. Additionally, DFDRL leverages a double deep Q-network (DDQN) to identify the best next-hop relay for each CH through interactions within the local environment. The reward function considers factors including residual energy deviation, distance to the Base Station (BS), and the number of neighboring CHs. Consequently, the inter-cluster energy efficiency and balance are greatly boosted. Furthermore, DFDRL incorporates a novel cluster maintenance mechanism to significantly reduce computation and message overheads, thereby further improving the network energy efficiency. Finally, extensive simulation experiments are conducted to evaluate the performance of DFDRL. The results demonstrates that DFDRL has superior network lifetime, lower energy consumption, and higher throughput compared to WOAD3QN-RP, EER-RL, HHOCFR and EEFRP in various scenarios.</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

A distributed cluster-based routing protocol using fuzzy logic and deep reinforcement learning for wireless sensor networks

  • 期刊论文

  • 发表日期：

    2025年8月19日

查看全文
  • Chuhang Wang,
  • Si Chen,
  • Huangshui Hu,
  • Xinji Fan

摘要

Cluster-based routing protocols play a crucial role in enhancing the applicability of wireless sensor networks (WSNs), which have been facing the challenge of maximizing network lifetime. Addressing this problem, this paper proposes a distributed, energy efficient and balanced clustering and routing protocol called DFDRL which integrates fuzzy logic with deep reinforcement learning (DRL). DFDRL selects optimal cluster heads (CHs) using fuzzy logic based on residual energy ratio, distance deviation, and node degree variance, which achieves intra-cluster energy efficiency and balance. Additionally, DFDRL leverages a double deep Q-network (DDQN) to identify the best next-hop relay for each CH through interactions within the local environment. The reward function considers factors including residual energy deviation, distance to the Base Station (BS), and the number of neighboring CHs. Consequently, the inter-cluster energy efficiency and balance are greatly boosted. Furthermore, DFDRL incorporates a novel cluster maintenance mechanism to significantly reduce computation and message overheads, thereby further improving the network energy efficiency. Finally, extensive simulation experiments are conducted to evaluate the performance of DFDRL. The results demonstrates that DFDRL has superior network lifetime, lower energy consumption, and higher throughput compared to WOAD3QN-RP, EER-RL, HHOCFR and EEFRP in various scenarios.