Scholarlink

<p>The Internet of Things (IoT) connects physical devices like sensors, actuators, and wearable technology to the Internet, enabling data gathering and sharing. The Internet of Medical Things (IoMT), a subset of IoT, focuses on healthcare, integrating devices to collect and transmit health-related data for processing and analysis. This paper introduces a Smart Internet of Medical Things (SIoMT) architecture that incorporates energy harvesting within Wireless Body Area Networks (WBAN). SIoMT integrates an online controller, RLCARE, which adapts application services in real-time to maximize user comfort while considering available harvested energy. We address the challenge of maximizing user comfort while minimizing energy consumption, a complex NP-hard problem due to the exponential increase in activation combinations with the number of services and the need to balance conflicting objectives. The proposed approach employs Reinforcement Learning (RL) techniques, particularly Pareto Q-learning, to manage these trade-offs effectively. RLCARE is modeled as a Markov Decision Process (MDP), enabling efficient decision-making that aligns with the problem’s sequential nature. We validated RLCARE through a proof of concept showing its ability to approximate the Pareto front of optimal solutions efficiently. The system’s performance was demonstrated through its adaptability to online variations in energy availability and application properties, as well as its efficiency in energy management by measuring saved energy. Comparisons with related methods reveal that RLCARE outperforms existing approaches, achieving up to 45% improvement in adaptation time and up to 30% more energy savings.</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

User comfort or sustainability? A multi-objective reinforcement learning strategy for the IoMT with harvesting energy

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Rania Mzid,
  • Bakhta Haouari,
  • Olfa Mosbahi

摘要

The Internet of Things (IoT) connects physical devices like sensors, actuators, and wearable technology to the Internet, enabling data gathering and sharing. The Internet of Medical Things (IoMT), a subset of IoT, focuses on healthcare, integrating devices to collect and transmit health-related data for processing and analysis. This paper introduces a Smart Internet of Medical Things (SIoMT) architecture that incorporates energy harvesting within Wireless Body Area Networks (WBAN). SIoMT integrates an online controller, RLCARE, which adapts application services in real-time to maximize user comfort while considering available harvested energy. We address the challenge of maximizing user comfort while minimizing energy consumption, a complex NP-hard problem due to the exponential increase in activation combinations with the number of services and the need to balance conflicting objectives. The proposed approach employs Reinforcement Learning (RL) techniques, particularly Pareto Q-learning, to manage these trade-offs effectively. RLCARE is modeled as a Markov Decision Process (MDP), enabling efficient decision-making that aligns with the problem’s sequential nature. We validated RLCARE through a proof of concept showing its ability to approximate the Pareto front of optimal solutions efficiently. The system’s performance was demonstrated through its adaptability to online variations in energy availability and application properties, as well as its efficiency in energy management by measuring saved energy. Comparisons with related methods reveal that RLCARE outperforms existing approaches, achieving up to 45% improvement in adaptation time and up to 30% more energy savings.