Pub Date : 2025-12-15DOI: 10.1109/tccn.2025.3644040
Jian Gu, Yin Wang, Wen Ji, Zhongxiang Wei, Jingjing Wang
{"title":"LLM-Based Dynamic Event-Triggered Communication for Multi-UAV Formation Control in Urban Environments","authors":"Jian Gu, Yin Wang, Wen Ji, Zhongxiang Wei, Jingjing Wang","doi":"10.1109/tccn.2025.3644040","DOIUrl":"https://doi.org/10.1109/tccn.2025.3644040","url":null,"abstract":"","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"77 1","pages":""},"PeriodicalIF":8.6,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145759615","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-15DOI: 10.1109/TCCN.2025.3644303
Jin Xie;Yunzhe Jiang;Ke Zhang;Fan Wu;Yin Zhang;Xiaoyan Huang;Shujiang Xu;Chau Yuen;Yan Zhang
In the forthcoming 6G paradigm, billions of endpoint devices will benefit from the widespread availability of network services. Given the unique characteristics of spectrum access across numerous and geographically dispersed devices, blockchain-based spectrum sharing (BSS) presents a compelling solution for enabling dynamic and decentralized spectrum allocation. However, the performance of blockchain, specifically the transaction throughput and block interval, directly impacts the efficiency of spectrum sharing. This aspect is often overlooked in current research. Furthermore, the decentralized nature of blockchain presents challenges for interference management due to the absence of centralized transmission power and channel coordination. To address these issues, we propose the Directed Acyclic graph and SHarding-based blockchain (DASH) for spectrum sharing, which improves transaction throughput while accounting for block interval effects in spectrum sharing scenarios. Additionally, we delve into a blockchain-assisted multi-agent deep reinforcement learning (MADRL) approach to tackle interference management in a decentralized manner. Finally, we evaluate our method taking into account the time delay associated with blockchain updates. The numerical results demonstrate the effectiveness of our proposed approach.
{"title":"High-Throughput DAG Blockchain for Efficient Spectrum Sharing in 6G Networks","authors":"Jin Xie;Yunzhe Jiang;Ke Zhang;Fan Wu;Yin Zhang;Xiaoyan Huang;Shujiang Xu;Chau Yuen;Yan Zhang","doi":"10.1109/TCCN.2025.3644303","DOIUrl":"10.1109/TCCN.2025.3644303","url":null,"abstract":"In the forthcoming 6G paradigm, billions of endpoint devices will benefit from the widespread availability of network services. Given the unique characteristics of spectrum access across numerous and geographically dispersed devices, blockchain-based spectrum sharing (BSS) presents a compelling solution for enabling dynamic and decentralized spectrum allocation. However, the performance of blockchain, specifically the transaction throughput and block interval, directly impacts the efficiency of spectrum sharing. This aspect is often overlooked in current research. Furthermore, the decentralized nature of blockchain presents challenges for interference management due to the absence of centralized transmission power and channel coordination. To address these issues, we propose the Directed Acyclic graph and SHarding-based blockchain (DASH) for spectrum sharing, which improves transaction throughput while accounting for block interval effects in spectrum sharing scenarios. Additionally, we delve into a blockchain-assisted multi-agent deep reinforcement learning (MADRL) approach to tackle interference management in a decentralized manner. Finally, we evaluate our method taking into account the time delay associated with blockchain updates. The numerical results demonstrate the effectiveness of our proposed approach.","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"12 ","pages":"4865-4881"},"PeriodicalIF":7.0,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145759614","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The rapid expansion of aerial vehicle applications in the low-altitude economy (LAE) requires reliable scene understanding to support safe and effective urban operations. However, existing 2D-based methods suffer from depth errors and occlusion, while direct 3D data transmission incurs unsustainable communication costs. Although semantic communication (SemCom) offers a promising alternative by transmitting only task-relevant features, its application to 3D scene understanding remains largely unexplored. To address these issues, we propose a task-oriented SemCom framework (TASC-3D), aiming to enhance 3D scene understanding in LAE networks. Specifically, TASC-3D integrates a hybrid encoder for effective scene representation, a diffusion model for enhanced noise resilience, and a large language model (LLM) to interpret high-level semantics into executable commands. A key challenge lies in the effective encoding of complex 3D environments, which fundamentally differs from 2D scenes due to the inherent structure and spatial complexity. To tackle this challenge, we introduce an object-level hybrid encoder that fuses geometric and visual semantics to provide a comprehensive and compact representation of 3D scenes. Then, an adaptive-rate channel denoising module is proposed for robust semantic transmission under fluctuating wireless conditions. Furthermore, to support multiple 3D perception tasks within a unified framework, we leverage an LLM to implement a unified multi-task formulation. Extensive experiments demonstrate that TASC-3D outperforms baseline methods in compression efficiency, transmission robustness, and downstream tasks accuracy, highlighting its potential for enabling practical 3D semantic communication in LAE aerial applications.
{"title":"LLM-Empowered Semantic Communication for Multi-Task 3D Scene Understanding in Low-Altitude Economy Networks","authors":"Jiawei Wang;Yang Tian;Junjie Li;Haofeng Sun;Hui Tian;Ping Zhang","doi":"10.1109/TCCN.2025.3642880","DOIUrl":"10.1109/TCCN.2025.3642880","url":null,"abstract":"The rapid expansion of aerial vehicle applications in the low-altitude economy (LAE) requires reliable scene understanding to support safe and effective urban operations. However, existing 2D-based methods suffer from depth errors and occlusion, while direct 3D data transmission incurs unsustainable communication costs. Although semantic communication (SemCom) offers a promising alternative by transmitting only task-relevant features, its application to 3D scene understanding remains largely unexplored. To address these issues, we propose a task-oriented SemCom framework (TASC-3D), aiming to enhance 3D scene understanding in LAE networks. Specifically, TASC-3D integrates a hybrid encoder for effective scene representation, a diffusion model for enhanced noise resilience, and a large language model (LLM) to interpret high-level semantics into executable commands. A key challenge lies in the effective encoding of complex 3D environments, which fundamentally differs from 2D scenes due to the inherent structure and spatial complexity. To tackle this challenge, we introduce an object-level hybrid encoder that fuses geometric and visual semantics to provide a comprehensive and compact representation of 3D scenes. Then, an adaptive-rate channel denoising module is proposed for robust semantic transmission under fluctuating wireless conditions. Furthermore, to support multiple 3D perception tasks within a unified framework, we leverage an LLM to implement a unified multi-task formulation. Extensive experiments demonstrate that TASC-3D outperforms baseline methods in compression efficiency, transmission robustness, and downstream tasks accuracy, highlighting its potential for enabling practical 3D semantic communication in LAE aerial applications.","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"12 ","pages":"4896-4910"},"PeriodicalIF":7.0,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145728968","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The personalized demands of diverse applications, the heterogeneity of space-air-ground integrated network (SAGIN) architectures, and the constrained and multidimensional nature of network resources all contribute to the significant issue of resource allocation and multi-vehicle load balancing within the realm of high-dynamic connected vehicles. To address this issue, this paper aims to propose a knowledge-driven artificial general intelligence approach that combines the advantages of both knowledge-driven and data-driven methods to handle complex task offloading problems in SAGIN-based vehicular networks. Specifically, the access network selection, computing resource and transmission power are jointly decided to minimize the system costs associated with imbalanced transmission and computation loads. To achieve this, the original problem is decomposed into two sub-problems of different temporal scales, considering knowledge about time granularity differences in executing various decisions. That is, the access network selection and computing resource allocation are handled at a large time scale, while power allocation is addressed at a smaller time scale. Within this two-timescale framework, a knowledge-driven deep reinforcement learning approach is proposed, further integrating model-based mathematical knowledge to obtain the closed-form power allocation at the small timescale and enforce the hard constraints at the large timescale via adding a safety layer. Numerical results show that the proposed knowledge-driven algorithm reduces system costs by 60% compared with the conventional one, while maintaining extremely low online inference latency even as the number of vehicles increases.
{"title":"Knowledge-Driven Two-Timescale Intelligent Task Offloading in SAGIN-Based Vehicular Networks","authors":"Ruijin Sun;Ge Qi;Lei Huang;Nan Cheng;Xiucheng Wang;Meng Qin","doi":"10.1109/TCCN.2025.3642318","DOIUrl":"10.1109/TCCN.2025.3642318","url":null,"abstract":"The personalized demands of diverse applications, the heterogeneity of space-air-ground integrated network (SAGIN) architectures, and the constrained and multidimensional nature of network resources all contribute to the significant issue of resource allocation and multi-vehicle load balancing within the realm of high-dynamic connected vehicles. To address this issue, this paper aims to propose a knowledge-driven artificial general intelligence approach that combines the advantages of both knowledge-driven and data-driven methods to handle complex task offloading problems in SAGIN-based vehicular networks. Specifically, the access network selection, computing resource and transmission power are jointly decided to minimize the system costs associated with imbalanced transmission and computation loads. To achieve this, the original problem is decomposed into two sub-problems of different temporal scales, considering knowledge about time granularity differences in executing various decisions. That is, the access network selection and computing resource allocation are handled at a large time scale, while power allocation is addressed at a smaller time scale. Within this two-timescale framework, a knowledge-driven deep reinforcement learning approach is proposed, further integrating model-based mathematical knowledge to obtain the closed-form power allocation at the small timescale and enforce the hard constraints at the large timescale via adding a safety layer. Numerical results show that the proposed knowledge-driven algorithm reduces system costs by 60% compared with the conventional one, while maintaining extremely low online inference latency even as the number of vehicles increases.","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"12 ","pages":"4986-5000"},"PeriodicalIF":7.0,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145728960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The intelligent manufacturing environment imposes extremely high requirements on real-time performance and accuracy. However, due to the dual constraints of communication and computing resources, meeting these requirements poses significant challenges. In this paper, we investigate how to achieve joint optimization of intelligence distribution and fine-tuning in the process of acquiring and applying intelligent models by agents. The framework efficiently provides intelligent models to the agents at a low cost while optimizing the intelligent models to ensure the overall performance of the system. Firstly, to achieve ubiquitous collaboration across computing resources driven by network awareness, we propose a multi-agent intelligent manufacturing architecture based on Computing Power Network (CPN). Secondly, to balance computing and communication resources and enhance model accuracy performance, we formulate a hierarchical optimization framework based on a dual spatial scale approach, combining intelligence distribution and fine-tuning. We further design a joint optimization algorithm based on the Differential Evolutionary (DE) framework and a synergy between Coalition Game Theory (CGT) and Federated Learning (FL) to effectively solve this problem. Finally, extensive simulation experiments validate the effectiveness and superiority of the proposed solution.
{"title":"Joint Intelligence Distribution and Fine-Tuning for Multi-Agent Intelligence Manufacturing","authors":"Renchao Xie;Anqi Zhou;Qinqin Tang;Tao Huang;Tianjiao Chen;Zehui Xiong","doi":"10.1109/TCCN.2025.3642816","DOIUrl":"10.1109/TCCN.2025.3642816","url":null,"abstract":"The intelligent manufacturing environment imposes extremely high requirements on real-time performance and accuracy. However, due to the dual constraints of communication and computing resources, meeting these requirements poses significant challenges. In this paper, we investigate how to achieve joint optimization of intelligence distribution and fine-tuning in the process of acquiring and applying intelligent models by agents. The framework efficiently provides intelligent models to the agents at a low cost while optimizing the intelligent models to ensure the overall performance of the system. Firstly, to achieve ubiquitous collaboration across computing resources driven by network awareness, we propose a multi-agent intelligent manufacturing architecture based on Computing Power Network (CPN). Secondly, to balance computing and communication resources and enhance model accuracy performance, we formulate a hierarchical optimization framework based on a dual spatial scale approach, combining intelligence distribution and fine-tuning. We further design a joint optimization algorithm based on the Differential Evolutionary (DE) framework and a synergy between Coalition Game Theory (CGT) and Federated Learning (FL) to effectively solve this problem. Finally, extensive simulation experiments validate the effectiveness and superiority of the proposed solution.","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"12 ","pages":"4850-4864"},"PeriodicalIF":7.0,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145728974","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-10DOI: 10.1109/tccn.2025.3642320
Yuqian Song, Jingli Zhou, Shudan Yu, Jun Liu
{"title":"QoS-Aware and Low-Cost Routing Optimization with Graph Reinforcement Learning in Hybrid Knowledge-Defined Networking","authors":"Yuqian Song, Jingli Zhou, Shudan Yu, Jun Liu","doi":"10.1109/tccn.2025.3642320","DOIUrl":"https://doi.org/10.1109/tccn.2025.3642320","url":null,"abstract":"","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"9 1","pages":""},"PeriodicalIF":8.6,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145717745","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-09DOI: 10.1109/TCCN.2025.3635940
{"title":"IEEE Communications Society Information","authors":"","doi":"10.1109/TCCN.2025.3635940","DOIUrl":"https://doi.org/10.1109/TCCN.2025.3635940","url":null,"abstract":"","PeriodicalId":13069,"journal":{"name":"IEEE Transactions on Cognitive Communications and Networking","volume":"11 6","pages":"C3-C3"},"PeriodicalIF":7.0,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11288934","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705900","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：