Scholarlink

<p>To address the challenging problem of collaborative optimization of communication delay and UAV load balancing in multi-Unmanned Aerial Vehicle (UAV)-assisted wireless rechargeable sensor networks, a dynamic threshold-enhanced diffusion proximal policy optimization algorithm (DTD-PPO) is proposed. Firstly, a multi-objective optimization model of multi-UAV-assisted WRSNs is constructed, and multi-dimensional constraints are incorporated to enhance the feasibility and practicality of the optimization solution. Secondly, a Markov Decision Process (MDP) framework is designed to balance the conflict between the dual objectives through dynamic weighting. To improve the exploration ability and training stability of the algorithm, the diffusion model is integrated into the PPO policy network, generating diversified actions through an adaptive noise-adding and denoising process. Additionally, a dynamic threshold strategy based on the normalized reward change rate is proposed to adjust the policy update magnitude in real-time. The effectiveness of our proposed algorithm is validated by using metrics of the data collection delay, UAV’s flight distance deviation and the energy efficiency. The simulation results verify the superiority and robustness of DTD-PPO algorithm compared to the other benchmark methods.</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

Dynamic threshold-enhanced diffusion PPO for multi-UAV collaborative optimization in wireless rechargeable sensor networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月4日

查看全文
  • Yalin Nie,
  • Zeyu Sun,
  • Yang Zhang

摘要

To address the challenging problem of collaborative optimization of communication delay and UAV load balancing in multi-Unmanned Aerial Vehicle (UAV)-assisted wireless rechargeable sensor networks, a dynamic threshold-enhanced diffusion proximal policy optimization algorithm (DTD-PPO) is proposed. Firstly, a multi-objective optimization model of multi-UAV-assisted WRSNs is constructed, and multi-dimensional constraints are incorporated to enhance the feasibility and practicality of the optimization solution. Secondly, a Markov Decision Process (MDP) framework is designed to balance the conflict between the dual objectives through dynamic weighting. To improve the exploration ability and training stability of the algorithm, the diffusion model is integrated into the PPO policy network, generating diversified actions through an adaptive noise-adding and denoising process. Additionally, a dynamic threshold strategy based on the normalized reward change rate is proposed to adjust the policy update magnitude in real-time. The effectiveness of our proposed algorithm is validated by using metrics of the data collection delay, UAV’s flight distance deviation and the energy efficiency. The simulation results verify the superiority and robustness of DTD-PPO algorithm compared to the other benchmark methods.