Scholarlink

<p>Unmanned aerial vehicle (UAV)-enabled Internet of Things (IoT) networks face a fundamental challenge in data collection when IoT devices are unevenly distributed across multiple clusters. As the UAV moves among these clusters, it encounters a dynamically varying communication environment, which often results in inefficient transmission energy consumption at the device side. While hybrid non-orthogonal multiple access (NOMA) provides greater flexibility in adapting resource allocation to heterogeneous device densities, it also introduces severe combinatorial complexity due to device pairing and successive interference cancellation (SIC) decoding-order design. This challenge is further exacerbated by the coupling of these discrete decisions with continuous transmit-power control and UAV trajectory optimization. To address this issue, we propose a hierarchical reinforcement learning (HRL)-based method that decomposes the original joint optimization problem into two coupled layers. The outer layer determines the UAV order by sequentially selecting the next cluster to visit, whereas the inner layer determines the UAV hovering position and the corresponding hybrid-NOMA resource allocation strategy within the selected cluster. Based on this decomposition, we formulate a two-tier decision model, in which the state transition and reward of the outer layer are dynamically generated according to the optimization result of the inner layer, thereby enabling collaborative optimization between inter-cluster trajectory and intra-cluster communication control. Simulation results show that the proposed method reduces transmission energy consumption by 25.2% on average and shortens the training time by 35.6% compared with benchmark schemes.</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

Energy-aware data collection using hierarchical reinforcement learning for UAV-enabled IoT networks

  • 期刊论文

  • 发表日期：

    2026年5月22日

查看全文
  • Xiaoli Han,
  • Bo Hu,
  • Renpeng Liu,
  • Tianle Wang

摘要

Unmanned aerial vehicle (UAV)-enabled Internet of Things (IoT) networks face a fundamental challenge in data collection when IoT devices are unevenly distributed across multiple clusters. As the UAV moves among these clusters, it encounters a dynamically varying communication environment, which often results in inefficient transmission energy consumption at the device side. While hybrid non-orthogonal multiple access (NOMA) provides greater flexibility in adapting resource allocation to heterogeneous device densities, it also introduces severe combinatorial complexity due to device pairing and successive interference cancellation (SIC) decoding-order design. This challenge is further exacerbated by the coupling of these discrete decisions with continuous transmit-power control and UAV trajectory optimization. To address this issue, we propose a hierarchical reinforcement learning (HRL)-based method that decomposes the original joint optimization problem into two coupled layers. The outer layer determines the UAV order by sequentially selecting the next cluster to visit, whereas the inner layer determines the UAV hovering position and the corresponding hybrid-NOMA resource allocation strategy within the selected cluster. Based on this decomposition, we formulate a two-tier decision model, in which the state transition and reward of the outer layer are dynamically generated according to the optimization result of the inner layer, thereby enabling collaborative optimization between inter-cluster trajectory and intra-cluster communication control. Simulation results show that the proposed method reduces transmission energy consumption by 25.2% on average and shortens the training time by 35.6% compared with benchmark schemes.