Research on KE-PPO-based Semantic Knowledge Alignment and Dynamic Adaptive Transmission for 6G Edge Agents

ZHUGE Bin; Hong Shiyu; Xu Yunhan; DONG Ligang; ZHANG Zitian

doi:10.11959/j.issn.1000-0801.DXKX260303

Chinese

您当前的位置：

首页 >

文章列表页 >

Research on KE-PPO-based Semantic Knowledge Alignment and Dynamic Adaptive Transmission for 6G Edge Agents

更新时间：2026-09-29

- Research on KE-PPO-based Semantic Knowledge Alignment and Dynamic Adaptive Transmission for 6G Edge Agents
- Telecommunications Science (2026)
- 作者机构：
  
  浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院），浙江杭州 310018
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(W2421086;61871468);Key Science and Technology Innovation Project of Zhejiang Province(2023R5211);Key R&D Program of Zhejiang Province(2025C02038);Project of Tongxiang General Artificial Intelligence Research Institute(TAGI2-B-2024-0014);Zhejiang Key Laboratory of New Network Standards and Application Technology(2013E10012)
- DOI：10.11959/j.issn.1000-0801.DXKX260303
  CLC： TP393
- Received：13 May 2026，
  
  Revised：2026-07-12，
  
  Accepted：02 September 2026，
- 稿件说明：
移动端阅览
ZHUGE Bin, Hong Shiyu, Xu Yunhan, et al. Research on KE-PPO-based Semantic Knowledge Alignment and Dynamic Adaptive Transmission for 6G Edge Agents[J/OL]. Telecommunications Science, 2026.
DOI：

ZHUGE Bin, Hong Shiyu, Xu Yunhan, et al. Research on KE-PPO-based Semantic Knowledge Alignment and Dynamic Adaptive Transmission for 6G Edge Agents[J/OL]. Telecommunications Science, 2026. DOI： 10.11959/j.issn.1000-0801.DXKX260303.

摘要

针对6G语义通信分布式协同中高昂信令开销与环境非平稳性难题，本文提出一种面向深层语义智能体的动态经验对齐与自适应传输机制。研究首先将边缘协同通信过程重构为单智能体部分可观察马尔可夫决策过程（POMDP），利用局部敏感哈希将高维语义特征投影至低维流形，构建轻量级动态语义知识库。随后，提出知识增强型近端策略优化（KE-PPO）算法，结合信息瓶颈理论重构联合奖励函数，并引入策略熵驱动的动态融合权重机制。基于真实视觉数据集与时变衰落信道的实验表明，该机制降低了全局交互负担，控制面交互量估算较传统分布式方案降低约两个数量级（在所设定的协议交互模型下理论降低约150倍）；在-10 dB极端信道环境下，推理准确率较DeepSC提升约16.5%；同时，边缘节点在线更新算力开销成功维持在亚毫秒级。结果证实，该机制在一定程度上抑制了非平稳环境下的策略震荡，在5个独立随机种子下的实验表明，其在保障实时性的同时，实现了推理精度与通信时延的有效折中。

Abstract

Addressing the challenges of prohibitive signaling overhead and environmental non-stationarity in distributed coordination for 6G semantic communications

this paper proposes a dynamic experience alignment and adaptive transmission mechanism for deep semantic agents. First

the edge collaborative communication process is reconstructed as a single-agent Partially Observable Markov Decision Process (POMDP). A Locality-Sensitive Hashing (LSH) operator is introduced to project high-dimensional semantic features onto low-dimensional manifolds

thereby constructing a lightweight dynamic semantic knowledge base . Subsequently

a Knowledge-Enhanced Proximal Policy Optimization (KE-PPO) algorithm is developed

which reconstructs the joint reward function based on Information Bottleneck (IB) theory and incorporates a policy entropy-driven dynamic fusion weight mechanism. Experiments based on real-world visual datasets and time-varying fading channels demonstrate that the mechanism significantly reduces the global interaction burden. The theoretical estimation of control plane signaling overhead is reduced by about two orders of magnitude compared to traditional distributed schemes (theoretically reduced by about 150 times under the assumed protocol interaction model)

and inference accuracy is improved by approximately 16.5% compared to DeepSC under extreme channel conditions at -10 dB SNR. Furthermore

the online update computational overhead at edge nodes is successfully maintained at a sub-millisecond level. Results confirm that the mechanism suppresses policy oscillations in non-stationary environments to a certain extent. Experiments under 5 independent random seeds show that

while ensuring real-time performance

it achieves an effective trade-off between inference accuracy and communication latency.

关键词

Keywords

references

STRINATI E C , BARBAROSSA S , GONZALEZ-JIMENEZ J L , et al . 6G networks: Beyond Shannon towards semantic and goal-oriented communications [J ] . Computer Networks , 2021 , 190 : 107930 .

ZHANG P , XU W , GAO H , et al . Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks [J ] . Engineering , 2022 , 8 : 60 - 73 .

SHAO J , MAO Y , ZHANG J . Learning task-oriented communication for edge inference: An information bottleneck approach [J ] . IEEE Journal on Selected Areas in Communications , 2022 , 40 ( 1 ): 197 - 211 .

GÜNDÜZ D , QIN Z , AGASOUNIS I , et al . Beyond transmitting bits: Context, semantics, and task-oriented communications [J ] . IEEE Journal on Selected Areas in Communications , 2022 , 41 ( 1 ): 5 - 41 .

XIE H , QIN Z , LI G Y , et al . Deep learning enabled semantic com-munication systems [J ] . IEEE Transactions on Signal Processing , 2021 , 69 : 2663 - 2679 .

YANG W , DU H , LIEW S C , et al . Semantic communications for future internet: Fundamentals, applications, and challenges [J ] . IEEE Communications Surveys & Tutorials , 2023 , 25 ( 1 ): 213 - 250 .

SHI G , XIAO Y , LI Y , et al . From semantic communication to se-mantic-aware networking: Model, architecture, and open problems [J ] . IEEE Communications Magazine , 2021 , 59 ( 8 ): 44 - 50 .

JI Y , WANG Y , ZHAO H , et al . Multi-agent reinforcement learning resources allocation method using dueling double deep Q-network in vehicular networks [J ] . IEEE Transactions on Vehicular Technology , 2023 , 72 ( 10 ): 13447 - 13460 .

DU H , LI Z , NIYATO D , et al . Enabling AI-Generated Content Ser-vices in Wireless Edge Networks [J ] . IEEE Wireless Communications , 2024 , 31 ( 3 ): 226 - 234 .

YANG Q , LUO X , et al . Joint Transmission and Understanding of Semantics with Edge Intelligence [C ] // IEEE International Conference on Communications (ICC) , 2023 : 1 - 6 .

JI Y , et al . Multi-Agent Reinforcement Learning for Covert Se-mantic Communications over Wireless Networks [C ] // 2023 IEEE In-ternational Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2023 : 1 - 5 .

WANG Y , CHEN M , LUO T , et al . Performance optimization for semantic communications: An attention-based reinforcement learning approach [J ] . IEEE Journal on Selected Areas in Communications , 2022 , 40 ( 9 ): 2598 - 2613 .

ZHOU F , LI Y , ZHANG X , et al . Cognitive semantic communica-tion systems driven by knowledge graph [J ] . IEEE Transactions on Wireless Communications , 2023 , 22 ( 3 ): 1234 - 1248 .

WANG Y , et al . Semantic Communication for Partial Observation Multi-agent Reinforcement Learning [C ] // 2023 IEEE Statistical Signal Processing Workshop (SSP) , 2023 : 1 - 5 .

QI Q , WANG J , MA Z , et al . Knowledge-enhanced semantic com-munication system with OFDM transmissions [J ] . Science China Infor-mation Sciences , 2023 , 66 ( 8 ): 182304 .

YU C , VELU A , WANG B , et al . The surprising effectiveness of PPO in cooperative multi-agent games [C ] // Advances in Neural Infor-mation Processing Systems (NeurIPS) , 2022 , 35 : 24611 - 24624 .

SHAO Y , LOURO C P , GUNDUZ D . Learning over-the-air digital semantic communication [J ] . IEEE Journal on Selected Areas in Communications , 2023 , 41 ( 8 ): 2490 - 2505 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Intelligent anti-jamming decision algorithm of bivariate frequency hopping pattern based on ET-PPO

长文本动态感知分块与语义嵌入协同优化算法

深度研究智能体的关键技术发展及在运营商的应用展望

TAPFormer: 面向LLM的工具调用性能预测

MO-GRPO算法驱动的AI推理动态调度方案

Related Author

Yibo CHEN

Zhijin ZHAO

Related Institution

No data

AI问答

Postal code：100079
Tel：（010）53879206 Email：tmw@bjxintong.com.cn
Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09082226号-64 京公网安备11010602201714号
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.

⁰