| 左路路* ** ***,杨哲铭*,郑清芳**,纪雯* **.面向低碳排放的边缘节点部署和资源分配联合优化方法[J].高技术通讯(中文),2026,36(8):812~822 |
| 面向低碳排放的边缘节点部署和资源分配联合优化方法 |
| Low-carbon-emissions oriented joint optimization of edge node deployment and resource allocation |
| |
| DOI：10. 3772 / j. issn. 1002 - 0470. 2026. 08. 004 |
| 中文关键词: 边缘服务器部署； 碳排放； 多跳传输； 资源分配； 马尔可夫优化决策 |
| 英文关键词: edge server deployment, carbon emission, multi-hop transmission, resource allocation, Markov optimization decision |
| 基金项目: |
| 作者 | 单位 | | 左路路* ** *** | （*中国科学院计算技术研究所处理器芯片全国重点实验室北京 100190）
（**鹏城实验室深圳 518055）
（***中国科学院大学北京 100049） | | 杨哲铭* | | | 郑清芳** | | | 纪雯* ** | |
|
| 摘要点击次数: 20 |
| 全文下载次数: 17 |
| 中文摘要: |
| 现有边缘部署方案主要依赖于无线接入点（access points，APs）的负载统计，未考虑实际环境下终端与APs的传输决策和任务资源分配对系统整体时延和碳排放的影响。为了应对这一问题，本文设计了基于任务时延、部署预算和有限服务器资源约束下的低碳部署模型，同时开发了联合边缘节点位置放置、任务传输和资源分配的三步迭代部署算法。首先，基于两步马尔可夫决策设计APs最大通信接入和最低计算资源限制下的任务多跳通信传输机制。然后，采用卡罗需-库恩-塔克（Karush-Kuhn-Tucker，KKT）条件和排序策略实现任务的资源分配。最后，依据贪婪匹配实现边缘服务器的位置部署。本文在公开的数据集上进行了不同算法的实验评估，结果表明，相比现有算法，所提算法可以将系统碳排放和任务平均时延降低17.0%以上。 |
| 英文摘要: |
| Existing edge deployment methods primarily rely on workload statistics of wireless access points (APs), overlooking the effects of transmission strategies between end devices and APs, as well as task resource allocation on delay and carbon emissions in real-world environments. To address this problem, this paper presents a low-carbon deployment model constrained by task delay, deployment budget, and limited server resources. Meanwhile, a three-step iterative deployment algorithm is developed that combines edge node deployment, task transmission, and resource allocation. Firstly, two-step Markov optimization is applied to formulate multi-hop transmission mechanisms for tasks surpassing maximum APs access or minimum computational resource limits. Then, resource allocation is implemented using Karush-Kuhn-Tucker (KKT) conditions combined with sorting strategy. Finally, edge server locations are determined via greedy matching. Massive experiments on public datasets reveal that our proposed method can reduce carbon emission and task average delay by more than 17.0% compared with the existing algorithms. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
