Open Radio Access Network(O-RAN)，有別於以往封閉的系統，主要促進生 態開放，甚至是開源。透過開放介面以及軟硬體，讓升級與維護成本大幅降低 並且解放以往受大廠綁定設備的情形讓更多的廠商有機會參與這個市場。同時 可藉由軟體開發，滿足不同客戶需求，增加營收項目以及提供更好的行動網路 服務。這觀點就像是在 App Store 或 Google Play 等商店上所締造的豐富產業鏈、 創造的新應用，消費者(行動網路服務商)可選購各家製造的手機(開放硬體)，可 安裝各種的應用程式(開放軟體)。
本研究將基於 OSC 提供之開源 O-RAN 軟體進行端到端的驗證，本次驗證 方式將會採用 Traffic Steering Use Case 作為應用情境。因目前無實際環境可供 驗證，我們將使用 ITRI 的 RAN Simulator 來作為驗證環境。
本論文也將開發兩支不同的 Traffic Steering xApp 來預測 Throughput 以及 HandOver 的時間點，並與傳統的 HandOver 方法進行比較。其中主要參考指標 有 Drop Rate 以及 HandOver 次數，經過實驗後證實使用 Regression 預測 Throughput 因為 Input 不夠多導致資訊不足無法正確預測準確流量值，使用分類 器後可將 Drop Rate 提升至接近 0%的表現。

In the past, mobile network equipment relied on major manufacturers to provide exclusive software and hardware, and operators had to rely on equipment manufacturers for equipment procurement and services, and could not provide customers with flexible/fast, customized, and differentiated services.
Open Radio Access Network (O-RAN) advocates an open interface and open software/hardware, so that hardware equipment can be easily purchased and upgraded, and through software creation, application services can be rapidly deployed to meet different customer needs and increase revenue projects. This point of view is like the rich industrial chain created on the open platform of iOS or Android and the creation of new applications. Consumers can buy mobile phones from various manufacturers and install various applications.
This research will perform end-to-end verification based on the open source O- RAN software provided by OSC. This verification method will use Traffic Steering Use Case as the application scenario. Since there is currently no actual environment for verification, we will use ITRI's RAN Simulator as the verification environment.
This paper will also develop two different Traffic Steering xApps to predict the Throughput and the time points of HandOver, and compare with the traditional HandOver method. The main reference indicators are the Drop Rate and the number of HandOvers. After experiments, it was confirmed that the use of Regression to predict the Throughput could not correctly predict the accurate traffic value due to insufficient information due to insufficient Input. After using the classifier, the Drop Rate can be increased to nearly 0% performance.

參考文獻

[1] "O-RAN Alliance," [Online]. Available: https://www.o-ran.org/.
[2] "Acumos AI," [Online]. Available: https://www.acumos.org/.
[3] Harish Kumar Dureppagari , Ujwal Dinesha , Raini Wu , Santosh Ganji , Woo-Hyun Ko, Srinivas Shakkottai, Dinesh Bharadia, "Realtime intelligent control for NextG cellular radio access networks," 2022.
[4] Marcin Dryjański, Łukasz Kułacz, Adrian Kliks, "Toward Modular and Flexible Open RAN Implementations in 6G Networks: Traffic Steering Use Case and O-RAN xApps," 2021.
[5] Ehab Ahmed Ibrahim, M.R.M. Rizk, Ehab F. Badran, "Study of LTE-R X2 handover based on A3 event algorithm using MATLAB," 2015.
[6] Tianqi Chen, Carlos Guestrin, "XGBoost: A Scalable Tree Boosting System," 2016.
[7] Akshay Balsubramani, Sanjoy Dasgupta, Yoav Freund, Shay Moran, "An adaptive nearest neighbor rule for classification," 2019.