面向语义通信网络的能效跨层优化
Cross-layer energy efficiency optimization for semantic communication networks
余开文 1樊仁和 1苟文龙 1俞传航 1武刚1
作者信息
- 1. 电子科技大学通信抗干扰全国重点实验室,成都 611731
- 折叠
摘要
语义通信关注传输信息的内在含义,通过语义提取可显著减少需要传输的数据量,提高通信效率,在未来智能设备通信场景中展现出巨大的潜力.然而,深度学习使能的语义编解码进一步加剧传统通信的能量消耗.针对该问题,本文提出一种联合跨层优化框架,并设计了一种语义能效指标来评估用户的体验质量和全局系统的能量损耗.将该优化过程建模为部分可观测的马尔可夫过程,联合优化物理层中的功率控制和语义层中的语义压缩配置:功率分配用于消除小区间干扰,语义压缩等级配置用于优化语义传输效率.仿真结果表明,所提框架和算法能够有效解决语义层和物理层的联合优化问题.
Abstract
Semantic communication focuses on the meaning of the transmitted information,which can significantly reduce the amount of data to be transmitted and improve the communication efficiency through semantic extraction,showing great potential in the future communication scenarios of smart devices.However,deep learning-enabled semantic codec further exacerbate the energy consumption of traditional communications.To address this problem,we propose a joint cross-layer optimization framework,and design a semantic energy efficiency metrics to evaluate the user's quality of experience and energy consumption of the global system.The optimization process is modeled as a partially observable Markov process.Jointly optimize power control in the physical layer and semantic compression allocation in the semantic layer:the power allocation is used to eliminate inter-cell interference,and the semantic compression level configuration is used to optimize the semantic transmission efficiency.Simulation results show that the proposed framework and algorithm can effectively solve the joint optimization problem of semantic and physical layers.
关键词
资源分配/语义通信/语义感知网络/能量效率/多智能体强化学习Key words
resource allocation/semantic communications/semantic-aware network/energy efficiency/multi-agent reinforcement learning引用本文复制引用
基金项目
中央高校基本科研业务费专项(2242022k60006)
出版年
2024
NETL
NSTL
万方数据