Scholarlink

<p>Wireless Sensor Networks (WSNs) are widely used across various domains but often face challenges in energy efficiency, latency, and reliability, particularly in dynamic and resource-constrained environments. This paper introduces Adaptive Multi-Objective Reinforcement Learning (AMORL), a novel method designed to address these challenges and improve routing efficiency. AMORL optimizes energy efficiency, latency, and reliability within the network, achieving superior performance even in adverse conditions. Its hybrid learning framework ensures scalability and consistent efficiency, making it well-suited for large and complex WSN setups. In extensive simulations across varying network sizes, AMORL was evaluated against Q-Learning, LEACH, and a Deep Q-Network (DQN) baseline. Results averaged over 30 independent trials with 95% confidence intervals demonstrate that AMORL significantly reduces energy consumption to 0.0589&#xa0;J, minimizes latency to 32.01&#xa0;ms, improves packet delivery to 30.76%, and increases throughput to 0.505 Mbps. Statistical tests confirm that these improvements are significant compared to all baselines. These results highlight the potential of AMORL to outperform traditional methods, providing a reliable and efficient solution for modern wireless sensor 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

Machine Learning-Driven Adaptive Routing for Efficient and Reliable Wireless Sensor Networks

  • 期刊论文

  • 发表日期：

    2026年2月27日

查看全文
  • Gummarekula Sattibabu,
  • A. Arunkumar Gudivada,
  • Ramprasad Ravula,
  • R. Anil Kumar,
  • Kapula Kalyani

摘要

Wireless Sensor Networks (WSNs) are widely used across various domains but often face challenges in energy efficiency, latency, and reliability, particularly in dynamic and resource-constrained environments. This paper introduces Adaptive Multi-Objective Reinforcement Learning (AMORL), a novel method designed to address these challenges and improve routing efficiency. AMORL optimizes energy efficiency, latency, and reliability within the network, achieving superior performance even in adverse conditions. Its hybrid learning framework ensures scalability and consistent efficiency, making it well-suited for large and complex WSN setups. In extensive simulations across varying network sizes, AMORL was evaluated against Q-Learning, LEACH, and a Deep Q-Network (DQN) baseline. Results averaged over 30 independent trials with 95% confidence intervals demonstrate that AMORL significantly reduces energy consumption to 0.0589 J, minimizes latency to 32.01 ms, improves packet delivery to 30.76%, and increases throughput to 0.505 Mbps. Statistical tests confirm that these improvements are significant compared to all baselines. These results highlight the potential of AMORL to outperform traditional methods, providing a reliable and efficient solution for modern wireless sensor networks.