Scholarlink

<p>With the advent of the 6&#xa0;G era, computing power is the new productivity in the era of the digital economy. In response to the growing global demand for computing power, the computing power network (CPN) came into being. The goal of the computing power network is to gather idle computing power, build a collaborative computing system of “cloud, edge, and end,” and improve the utilization rate of computing power resources in the whole network. In traditional cellular-based mobile edge computing, due to signal attenuation and interference between users, users at the edge of the cell have a great impact on the use of computing resources. UAV-assisted resource management is critical to developing sixth-generation networks, where aerial base stations have a higher chance of establishing line-of-sight connections with users on the ground due to their greater coverage. How to share computing resources more efficiently and how to optimize resource allocation are urgent problems to be solved. To address these issues, this paper proposes a new user-centric computing power network architecture that considers unmanned aerial vehicles (UAVs) as access points with offloading capabilities and models the joint resource management process in user-centered UAV-assisted computing power network architecture (UCUAV-CPN) as a partially observable Markov decision process. A decentralized joint optimization scheme based on multi-agent deep reinforcement learning and convex optimization is proposed. The experimental results show that the scheme proposed in UCUAV-CPN can significantly reduce the average total delay by up to 83.58%, and the scheme proposed in this paper can also significantly improve the uplink transmission rate by up to 131.69%.</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

Task offloading and computing resource allocation of the joint UAV with computing power network

  • 期刊论文

  • 发表日期：

    2025年5月25日

查看全文
  • JunFei Li,
  • QingTao Zeng,
  • LiKun Lu,
  • ErQing Zhang,
  • AnPing Xu,
  • WenJing Li

摘要

With the advent of the 6 G era, computing power is the new productivity in the era of the digital economy. In response to the growing global demand for computing power, the computing power network (CPN) came into being. The goal of the computing power network is to gather idle computing power, build a collaborative computing system of “cloud, edge, and end,” and improve the utilization rate of computing power resources in the whole network. In traditional cellular-based mobile edge computing, due to signal attenuation and interference between users, users at the edge of the cell have a great impact on the use of computing resources. UAV-assisted resource management is critical to developing sixth-generation networks, where aerial base stations have a higher chance of establishing line-of-sight connections with users on the ground due to their greater coverage. How to share computing resources more efficiently and how to optimize resource allocation are urgent problems to be solved. To address these issues, this paper proposes a new user-centric computing power network architecture that considers unmanned aerial vehicles (UAVs) as access points with offloading capabilities and models the joint resource management process in user-centered UAV-assisted computing power network architecture (UCUAV-CPN) as a partially observable Markov decision process. A decentralized joint optimization scheme based on multi-agent deep reinforcement learning and convex optimization is proposed. The experimental results show that the scheme proposed in UCUAV-CPN can significantly reduce the average total delay by up to 83.58%, and the scheme proposed in this paper can also significantly improve the uplink transmission rate by up to 131.69%.