| 162 | 0 | 41 |
| 下载次数 | 被引频次 | 阅读次数 |
工业物联网(Industrial Internet of Things, IIoT)场景中，AI与边缘计算的深度融合正在逐渐兴起。然而，IIoT设备的通信和计算资源有限，且这些设备产生的数据通常隐私且高度异构，如何在确保数据安全的同时加速模型训练与推理，仍是一个开放的问题。联邦学习(Federated Learning, FL)作为一种新兴的隐私保护框架，在隐私保护和数据安全方面具有巨大潜力，但受限于资源有限的设备，其计算与通信效率的提升仍面临挑战。为解决上述问题，提出了FedEdgent框架，该框架基于深度神经网络(Deep Neural Network, DNN)的分割，构建了一个端边云协同系统。FedEdgent通过深度强化学习(Deep Reinforcement Learning, DRL)实现动态FL任务卸载策略，提升了端边云的协作训练效率。结合模型压缩技术，FedEdgent框架选择对全局FL模型贡献较大的DNN层进行优化，减少通信开销并提高通信效率。实验结果表明，相较于集中式FL,FedEdgent在保持相似准确率的同时，训练时间减少了约60%,上传参数量平均减少了25%。
Abstract：The deep integration of AI and edge computing emerges in the Industrial Internet of Things(IIoT) scenarios. However, IIoT devices have limited communication and computational resources, and data from these devices are usually highly heterogeneous and private, so how to accelerate training and inference while ensuring data security is still an open problem. Federated Learning(FL), has great advantages and potentials in privacy protection and data security, while remains the research problem with the efficient computational and communication due to the bottleneck of resource-constrained devices. This paper proposes FedEdgent framework, a hybrid device-edge-cloud synergy framework for accelerating FL based on the Deep Neural Network(DNN) partitioning approach, and implements a dynamic FL task offloading strategy through Deep Reinforcement Learning(DRL) to improve the training efficiency through device-edge-cloud collaboration. Combined with model compression, FedEdgent selects DNN layers with higher contributions to optimize the global FL model, which reduces the communication traffics, and improves the communication efficiency. Experimental results show that compared with the centralized FL, FedEdgent reduces the training time by about 60% and the amount of uploaded parameters by 25% on average while remaining comparable accuracy.
[1] VINUEZA-NARANJO P G,CHICAIZA J,RUMIPAMBA-ZAMBRANO R.Fog Computing Technology Research:A Retrospective Overview and Bibliometric Analysis[J].ACM Computing Surveys,2024,57(4):81.1-81.32.
[2] NECHIBVUTE A,MAFUKIDZE H D.Integration of SCADA and Industrial IoT:Opportunities and Challenges[J].IETE Technical Review,2024,41(3):312-325.
[3] QI Q S,XU Z Y,RANI P.Big Data Analytics Challenges to Implementing the Intelligent Industrial Internet of Things (IIoT) Systems in Sustainable Manufacturing Operations[J].Technological Forecasting and Social Change,2023,190:122401.1-122401.15.
[4] WU X,LIU Y,TIAN J,et al.Privacy-preserving Trust Management Method Based on Blockchain for Cross-domain Industrial IoT[J].Knowledge-Based Systems,2024,283(11):1.1-1.18.
[5] YU K P,TAN L,ALOQAILY M,et al.Blockchain-enhanced Data Sharing with Traceable and Direct Revocation in IIoT[J].IEEE Transactions on Industrial Informatics,2021,17(11):7669-7678.
[6] GUO Y T,LIU F,CAI Z P,et al.FEEL:A Federated Edge Learning System for Efficient and Privacy-preserving Mobile Healthcare[C]//Proceedings of the 49th International Conference on Parallel Processing.New York:ACM,2020:1-11.
[7] TANG B,HU B,QU Z H,et al.Optimized Power Control for Privacy-preserving Over-the-Air Federated Edge Learning with Device Sampling[J].IEEE Internet of Things Journal,2024,11(17):29157-29173.
[8] ZHANG C,ZHANG W J,WU Q,et al.Distributed Deep Reinforcement Learning Based Gradient Quantization for Federated Learning Enabled Vehicle Edge Computing[J].IEEE Internet of Things Journal,2025,12(5):4899-4913.
[9] LING X L,CHI W C,ZHANG J J,et al.Federated Learning Convergence Optimization for Energy-limited and Social-aware Edge Nodes[J].IEEE Access,2024,12:107844-107854.
[10] LIU X H,DONG X H,JIA N,et al.Federated Learning-oriented Edge Computing Framework for the IIoT[J].Sensors,2024,24(13):4182.
[11] RAFIQ A,WEI M,WANG P,et al.Delay Aware 6TiSCH IIoT Networks for Energy Efficient Data Transmission by Adopting Federated Learning and Edge Computing[J].IEEE Transactions on Consumer Electronics,2024,70(3):5911-5928.
[12] LO S K,LU Q G,WANG C,et al.A Systematic Literature Review on Federated Machine Learning:From a Software Engineering Perspective[J].ACM Computing Surveys,2021,54(5):95.1-95.39.
[13] LI E,ZHOU Z,CHEN X.Edge Intelligence:On-demand Deep Learning Model Co-inference with Device-edge Synergy[C]//Proceedings of the 2018 Workshop on Mobile Edge Communications.New York:ACM,2018:31-36.
[14] KANG Y P,HAUSWALD J,GAO C,et al.Neurosurgeon:Collaborative Intelligence Between the Cloud and Mobile Edge[J].ACM SIGARCH Computer Architecture News,2017,45(1):615-629.
[15] ESHRATIFAR A E,ABRISHAMI M S,PEDRAM M.JointDNN:An Efficient Training and Inference Engine for Intelligent Mobile Cloud Computing Services[J].IEEE Transactions on Mobile Computing,2019,20(2):565-576.
[16] ZHU G X,DU Y Q,GüNDüZ D,et al.One-bit Over-the-air Aggregation for Communication-efficient Federated Edge Learning:Design and Convergence Analysis[J].IEEE Transactions on Wireless Communications,2020,20(3):2120-2135.
[17] REN J K,YU G D,DING G Y.Accelerating DNN Training in Wireless Federated Edge Learning Systems[J].IEEE Journal on Selected Areas in Communications,2021,39(1):219-232.
[18] CAO S H,ZHANG H Q,WEN T,et al.FedQMIX:Communication-efficient Federated Learning via Multi-agent Reinforcement Learning[J].High-Confidence Computing,2024,4(2):100179.
[19] QIANG X K,HU Y,CHANG Z,et al.Importance-aware Data Selection and Resource Allocation for Hierarchical Federated Edge Learning[J].Future Generation Computer Systems,2024,154:35-44.
[20] HAPA C,ARACHCHIGE P C M,CAMTEPE S,et al.Splitfed:When Federated Learning Meets Split Learning[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Palo Alto:AAAI Press,2022:8485-8493.
[21] HE X,WEI J,YI Z H,et al.Optimal Power Flow Calculation Based on SplitNN-DNN[C]//Journal of Physics:Conference Series.Bristol:IOP Publishing,2024:012008.
[22] TURINA V,ZHANG Z,ESPOSITO F,MATTA I.Federated or Split?A Performance and Privacy Analysis of Hybrid Split and Federated Learning Architectures[C]//2021 IEEE 14th International Conference on Cloud Computing (CLOUD).Chicago:IEEE,2021:250-260.
[23] WU D,ULLAH R,HARVEY P,et al.FedAdapt:Adaptive Offloading for IoT Devices in Federated Learning[J].IEEE Internet of Things Journal,2022,9(21):20889-20901.
[24] BURD T D,BRODERSEN R W.Processor Design for Portable Systems[J].Journal of VLSI Signal Processing Systems for Signal,Image and Video Technology,1996,13(2):203-221.
[25] DINH C T,TRAN N H,NGUYEN M N H,et al.Federated Learning Over Wireless Networks:Convergence Analysisand Resource Allocation[J].IEEE/ACM Transactions on Networking,2021,29(1):398-409.
[26] MNIH V,KAVUKCUOGLU K,SILVER D,et al.Human-level Control Through Deep Reinforcement Learning[J].Nature,2015,518(7540):529-533.
基本信息:
中图分类号:TP309;TP181
引用信息:
[1]曹绍华,杨雁升,陈辉,等.FedEdgent:基于DNN分割的端边云协同联邦学习加速框架[J].无线电通信技术,2025,51(06):1306-1316.
基金信息:
山东省自然科学基金(ZR2024MF093)~~
2025-08-28
2025-08-28
2025-08-28