摘要
The evolution of open architectures for Radio Ac-cess Networks (RANs) is revolutionizing network management and optimization. This transformation, fostered by O-RAN, expedites data acquisition and examination by exploiting newly established open interfaces. Moreover, it has led to the rise of near real-time RAN Intelligent Controllers (RICs), instigating a wave of AI-driven applications, or xApps, that employ Arti-ficial Intelligence (AI)/Machine Learning (ML) methods. Nev-ertheless, deploying xApps as centralized applications presents substantial challenges, such as handling vast data transactions, potential delays, and security vulnerabilities, which are notably prominent within the multifaceted, decentralized, multivendor, and trustless nature of open networks. To alleviate these predicaments, a transition from centralized apps operating in near real-time to distributed real-time apps is imperative for augmented security and efficiency. This paper addresses these complexities by introducing an open platform that integrates a federated reinforcement learning algorithm to operate as distributed Apps (dApps) within the next-generation O-RAN architecture. We present evaluation results in a specific test environment.
学科与关键词
- Intelligent O-RAN Beyond 5G: Architecture, Use Cases, Challenges, and OpportunitiesMarinova, Simona*; Leon-Garcia, AlbertoIEEE Access, 2024, 12: 27088-27114.被引 4
- ColO-RAN: Developing Machine Learning-Based xApps for Open RAN Closed-Loop Control on Programmable Experimental PlatformsPolese, Michele*; Bonati, Leonardo; Doro, Salvatore; Basagni, Stefano; Melodia, TommasoIEEE Transactions on Mobile Computing, 2023, 22(10): 5787-5800.被引 50
- Accelerating Reinforcement Learning via Predictive Policy Transfer in 6G RAN SlicingNagib, Ahmad M.*; Abou-Zeid, Hatem; Hassanein, Hossam S.IEEE Transactions on Network and Service Management, 2023, 20(2): 1170-1183.被引 9
