Scholarlink

<p>In multi-UAV cooperative video streaming, the high mobility of UAVs and complex environmental obstructions degrade wireless communication, causing challenges for computility resources allocation. Existing studies on optimizing communication with Reconfigurable Intelligent Surfaces (RIS) often rely on stepwise or alternating optimization methods, which struggle to capture the coupling between RIS and resources. Furthermore, traditional single-layer Deep Reinforcement Learning (DRL) frameworks fail to handle the high-dimensional mixed action space composed of discrete phase and continuous resource allocation. To address these challenges, a DRL-based and RIS-assisted Hierarchical Soft Actor-Critic (HSAC) method is proposed, which decouples the action space into two subproblems. First, the upper-layer network makes optimal discrete phase-shift decisions with RIS to enhance the signal quality of key links. Then, the lower-layer agent builds a Markov Decision Process (MDP) for the resource allocation problem, focusing on the continuous and fine-grained allocation of computility resources based on the upper-layer decisions. The two layers collaborate through state awareness and a shared global reward to jointly optimize the Age of Information (AoI), transmission success probability, and video quality bitrate. Simulation results demonstrate that the HSAC outperforms baselines in terms of convergence speed, cumulative reward, and performance metrics, showcasing superior performance.</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

RIS-assisted computility networks resource allocation for multi-UAV cooperative video tasks

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Yin Yin,
  • Ningjiang Chen,
  • Guojun Gan

摘要

In multi-UAV cooperative video streaming, the high mobility of UAVs and complex environmental obstructions degrade wireless communication, causing challenges for computility resources allocation. Existing studies on optimizing communication with Reconfigurable Intelligent Surfaces (RIS) often rely on stepwise or alternating optimization methods, which struggle to capture the coupling between RIS and resources. Furthermore, traditional single-layer Deep Reinforcement Learning (DRL) frameworks fail to handle the high-dimensional mixed action space composed of discrete phase and continuous resource allocation. To address these challenges, a DRL-based and RIS-assisted Hierarchical Soft Actor-Critic (HSAC) method is proposed, which decouples the action space into two subproblems. First, the upper-layer network makes optimal discrete phase-shift decisions with RIS to enhance the signal quality of key links. Then, the lower-layer agent builds a Markov Decision Process (MDP) for the resource allocation problem, focusing on the continuous and fine-grained allocation of computility resources based on the upper-layer decisions. The two layers collaborate through state awareness and a shared global reward to jointly optimize the Age of Information (AoI), transmission success probability, and video quality bitrate. Simulation results demonstrate that the HSAC outperforms baselines in terms of convergence speed, cumulative reward, and performance metrics, showcasing superior performance.