Scholarlink

<p>Base station (BS) positioning remains a critical challenge in energy-efficient wireless sensor networks, as communication distance directly influences network lifetime. To address this challenge, federated learning (FL) is employed to optimize base station placement and sensor node sleep scheduling in a distributed manner. By exchanging only model parameters rather than cluster-state information, FL reduces the need for transmitting large amounts of network data while enabling collaborative learning across clusters. A single FL model coordinates both BS positioning and sleep scheduling decisions, allowing the network to adapt to changing conditions while improving energy efficiency. A neural network model is designed to process cluster-level features, including node locations, residual energy distribution, active node counts, cluster statistics, and sleep patterns, to simultaneously determine BS positions and sleep scheduling decisions. The framework further integrates energy trend prediction and coverage-awareness mechanisms to reduce redundant node activity while maintaining sensing coverage. Each cluster operates as an independent federated client, enabling local model training without sharing internal data. Model training is guided by BS positioning error while promoting energy efficiency, balanced load distribution, and adequate coverage. To ensure smooth and realistic BS mobility, trajectory smoothing and adjustment techniques are applied. Simulation results demonstrate that the proposed approach achieves lifetime improvements of 56.9% over KDL, 51.1% over E-LEACHFLS, 143.5% over E-LEACH and 45.8% over SEPWMBS-3L.</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

Federated Learning for Base Station Positioning and Sleep Scheduling in Wireless Sensor Networks

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Thierry Taning Longla,
  • Tansal Gucluoglu

摘要

Base station (BS) positioning remains a critical challenge in energy-efficient wireless sensor networks, as communication distance directly influences network lifetime. To address this challenge, federated learning (FL) is employed to optimize base station placement and sensor node sleep scheduling in a distributed manner. By exchanging only model parameters rather than cluster-state information, FL reduces the need for transmitting large amounts of network data while enabling collaborative learning across clusters. A single FL model coordinates both BS positioning and sleep scheduling decisions, allowing the network to adapt to changing conditions while improving energy efficiency. A neural network model is designed to process cluster-level features, including node locations, residual energy distribution, active node counts, cluster statistics, and sleep patterns, to simultaneously determine BS positions and sleep scheduling decisions. The framework further integrates energy trend prediction and coverage-awareness mechanisms to reduce redundant node activity while maintaining sensing coverage. Each cluster operates as an independent federated client, enabling local model training without sharing internal data. Model training is guided by BS positioning error while promoting energy efficiency, balanced load distribution, and adequate coverage. To ensure smooth and realistic BS mobility, trajectory smoothing and adjustment techniques are applied. Simulation results demonstrate that the proposed approach achieves lifetime improvements of 56.9% over KDL, 51.1% over E-LEACHFLS, 143.5% over E-LEACH and 45.8% over SEPWMBS-3L.