OpenRAN Gym: An Open Toolbox for Data Collection and Experimentation with AI in O-RAN

tags: 5G Reading

Date : 2022-11-04

Metadata

paper link
Bonati, L., Polese, M., D’Oro, S., Basagni, S., & Melodia, T. (2022, April). OpenRAN Gym: An Open Toolbox for Data Collection and Experimentation with AI in O-RAN. In 2022 IEEE Wireless Communications and Networking Conference (WCNC) (pp. 518-523). IEEE.

Take away

OpenRAN Gym is an end-to-end test environment for the O-RAN related experiments. It is both beneficial for academy and industrial use for decreasing the costs on operating the experiments and control the variables in experiments, etc.

Summary

The proposed OpenRAN Gym integrates several O-RAN components service projects into an end-to-end data-driven experimentation platform. Including Colosseum as the dataset, ColO-RAN as the O-RAN control architecture, SCOPE as the RAN APIs and srsRAN(or OpenAirInterface; OAI) as the softwarized RAN.

Note

• Overall architecture in OpenRAN Gym
  

• Data Collection and Control Framework - SCOPE
  paper link
  SCOPE利用現有虛擬化RAN平台srsRAN，加上額外的網路切片(network slicing)及規範排程(scheduling policies)等功能，達成能夠及時更改RAN相關設定的模擬能力。另外本專案與現有網路環境相關資料庫Colosseum有很高的貼合度，除了資料欄位能夠有對應的接口，例如QoS, scheduling policies, slicing resources等，以及透過E2 Termination能以OSC-DU接上Near-RT RIC平台，能夠加速應用的開發。
  

  
  

• O-RAN Control Architecture - ColO-RAN
  paper link
  ColO-RAN以輕量化的形式實作了OSC Near-RT RIC平台，透過部屬於Docker的環境，提供了O-RAN E2 Termination, E2 Manager, E2 Routing Manager, Redis DB, 和xApp Framework等必要的元件。部署完成後便能夠透過E2 Interface將ColO-RAN串連至SCOPE RAN，透過ColO-RAN上部署的xApp收到從RAN上傳的KPMs資料提供後續分析。
  

  
  

• Testbed - Colosseum
  paper link
  Colosseum是一個龐大的網路環境模擬專案，提供許多學術單位對網路環境模擬的需求，特別在網路環境中部署訓練AI模型的應用上在未來有更高的使用程度，原因為在網路環境中收集資料及訓練需要龐大的成本，若是能夠在模擬的環境中進行訓練調整的模型，除了降低成本外，也能夠對網路環境的各項變數控制有更高的掌握程度。Colosseum提供128 Standard Radio Nodes (SRNs), Massive Channel Emulator (MCHEM), Radio Frequency (RF) scenario server以及Traffic Generator (TGEN)。