A Digital Twin Based Learning Architecture for Resource Allocation in O-RAN

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/95365

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/95365

Title:	A Digital Twin Based Learning Architecture for Resource Allocation in O-RAN
Authors:	林彥澄;Lin, Yen-Chen
Contributors:	通訊工程學系
Keywords:	開放式無線網路;排程演算法;無線基地台;Open RAN;Celluar;6G;Scheduling;ns-3
Date:	2024-08-15
Issue Date:	2024-10-09 16:42:55 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著網路環境的持續發展，當前網路性能已不能滿足期望，需要進一步改進。O-RAN 引入 AI/ML workflow，旨在實現網路優化、預測性維護、智能流量管理、安全檢測和服務保證。將 AI/ML 整合到 O-RAN 中，運營商可以創建更高效、可靠和智能的網路，提供更優質的服務，降低成本，並適應不斷變化的需求。為解決對 AI/ML 模型信任度的問題，可採取包括選擇透明度高的模型、進行嚴格測試、遵守標準規格，以及進行跨領域協作。在虛擬環境中部署智能控制方法，如 near-RT RIC 的 xApp，可驗證演算法可行性並最小化錯誤決策。通過結合 AI/ML 技術的培訓環境，運營商能夠建立自動化的應用開發流程，同時降低與 AI/ML 部署相關的風險，從而實現網路性能的全面提升。;With the continuous development of the network environment, the current net-work performance has fallen short of expectations and there is a need for further improvement. Hence, O-RAN introduces AI/ML workflows to achieve network optimization, predictive maintenance, intelligent traffic management, security and anomaly detection, and customer experience. By integrating AI/ML workflows into O-RAN, operators can leverage the power of data-driven decision-making, automation, and optimization to create more efficient, reliable, and intelligent mobile networks. This enables them to deliver better service quality, reduce costs, and adapt to the evolving needs of mobile communication systems. To address concerns regarding trust in AI/ML models where network operators have limited control, several measures can be taken. Firstly, selecting models that offer transparency and explainability ensures operators can understand and interpret the decision-making process. Additionally, rigorous testing and validation in various scenarios, including simulated and real-world environments, help evaluate performance and reliability. Adhering to industry standards, regulations, and implementing security measures and privacy compliance also fosters trust. Collaborative development and peer reviews involving experts from different domains provide external validation. When comparing AI/ML solutions, considering vendor reputation, track record, performance metrics, and customer reviews is crucial. Lastly, to ensure real network performance, deploying intelligent control methods in virtual environments, such as near-RT RIC’s xApp, can verify algorithm feasibility and minimize erroneous decisions. By leveraging training environments that combine expert knowledge with AI/ML techniques, operators can establish fully automated app development processes and mitigate risks associated with AI/ML deployments.
Appears in Collections:	[Graduate Institute of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	5	View/Open

社群 sharing

Loading...