$30 off During Our Annual Pro Sale. View Details »

Case Studied on Federated learning Deployment in O-RAN

Case Studied on Federated learning Deployment in O-RAN

簡報下載 (Creative Commons BY 4.0):https://reurl.cc/eXejzj
Case Studied on Federated learning Deployment in O-RAN
O-RAN 中部署聯邦學習的案例研究
專題研究計畫-進度討論會議
日期:2023/05/11
國立清華大學 資訊工程學系 邱德泉教授 ACE LAB 蔡秀吉
[email protected]

參考資料來源:
https://thc1006.blogspot.com/2023/01/near-rt-ric.html
O-RAN.WG2.AIML-v01.03
Deep Learning for B5G Open Radio Access Network: Evolution, Survey, Case Studies, and Challenges: http://dx.doi.org/10.1109/OJCOMS.2022.3146618
https://thc1006.blogspot.com/2023/04/blog-post.html
https://wiki.o-ran-sc.org/display/OAM/OAM+Architecture

蔡秀吉

May 11, 2023
Tweet

More Decks by 蔡秀吉

Other Decks in Research

Transcript

  1. Case Studied on Federated learning
    Deployment in O-RAN
    2023/05/11

    View Slide

  2. View Slide

  3. Reference: O-RAN.WG1.OAD-R003-v08.00

    View Slide

  4. Reference: O-RAN.WG1.OAD-R003-v08.00

    View Slide

  5. O-RAN RAN Intelligent Controller (RIC)


    View Slide

  6. Open fronthaul

    View Slide

  7. O1
    O2 A1 OFH
    (M)
    O1
    O1
    O1
    O-Cloud
    O-DU O-RU O-RU
    Near-RT RIC
    Near-RT RIC
    O-CU-CP
    O-CU-UP
    Non-RT RIC
    O1
    rAPP
    SMO



    Reference: https://thc1006.blogspot.com/2023/04/blog-post.html

    View Slide

  8. Case Studied on FL Deployment in O-RAN


    Reference: O-RAN.WG2.AIML-v01.03

    View Slide

  9. FL is suitable for O-RAN, is also based on a Disaggregated
    and Distributed split architecture (O-DU and O-RU).
    O-RU Subnet
    O-DU
    Single O-RU
    O-CU-CP
    O-CU-UP
    Reference:Deep Learning for B5G Open Radio Access Network: Evolution, Survey, Case Studies, and Challenges

    View Slide

  10. O-RU Subnet
    O-DU
    Single O-RU
    Local Model
    Local Model

    View Slide

  11. Open-Fronthaul (O-FH) Split Option 7.2x 規格
    FL Split Option
    Fronthaul
    D to A | A to D
    RU DU

    View Slide

  12. Local Model
    Local Model
    Local Model
    O-DU
    O-CU-CP
    O-CU-UP
    F1-C
    F1-U
    O1
    O-FH
    Models Aggregation

    View Slide

  13. Train a Global Model
    Deploy Global model in
    “Related” Near RT RIC xApps
    Models Aggregation
    Deploy via A1
    Reference: O-RAN.WG2.AIML-v01.03

    View Slide

  14. • 因為應用情境 (scenario) 的不同, Near-RT RIC 會需要運行多個應用程式(xApps),
    進行資料監控,並對 RAN 功能(Function)的參數進行調整,才能達成應用情境
    的需求;Near-RT RIC 的控制決策,仍須根據 Non-RT RIC 提供的 Policy 與指令
    等…進行 RAN 控制的調整優化。
    Introduction to xApps
    Reference:https://thc1006.blogspot.com/2023/01/near-rt-ric.html

    View Slide

  15. Local Model
    Local Model
    Local Model
    O-DU
    O-CU-CP
    O-CU-UP
    F1-C
    F1-U
    O-FH
    (Network Functions,NFs)
    (O-CU、O-DU、O-RU)

    View Slide

  16. 將來自 NF 的數據進行儲存:
    Reference: https://wiki.o-ran-sc.org/display/OAM/OAM+Architecture

    View Slide