Radio access network (RAN) slicing is one of the key technologies in 5G and beyond mobile networks, where multiple logical subnets, i.e., RAN slices, are allowed to run on top of the same physical infrastructure so as to provide slice-specific services. Due to the dynamic environments of wireless networks and the diverse requirements of RAN slices, inter-slice radio resource management (IS-RRM) has become a highly challenging task in RAN slicing. In this paper, we propose a novel online convex optimization (OCO) framework for IS-RRM, which directly learns the instant resource allocation from the data revealed by previous allocations, such that sophisticated modeling and parameterization can be avoided in highly complicated and dynamic wireless environments. Specifically, an online IS-RRM scheme that employs multiple expert-algorithms running in parallel is proposed to keep track of the environmental changes and adjust the resource allocation accordingly. Both theoretical analysis and simulation results show that our proposed scheme can guarantee long-term performance comparable to the optimal strategies given in hindsight.

中文翻译：

---

用于高效、稳健的切片间无线电资源管理的在线凸优化

无线接入网 (RAN) 切片是 5G 及以后移动网络的关键技术之一，其中允许多个逻辑子网，即 RAN 切片运行在同一物理基础设施之上，以提供切片特定的服务。由于无线网络的动态环境和 RAN 切片的多样化需求，切片间无线电资源管理 (IS-RRM) 已成为 RAN 切片中极具挑战性的任务。在本文中，我们为 IS-RRM 提出了一种新颖的在线凸优化 (OCO) 框架，该框架直接从先前分配所揭示的数据中学习即时资源分配，从而可以避免在高度复杂和动态的无线网络中进行复杂的建模和参数化。环境。具体来说，提出了一种采用多个并行运行的专家算法的在线 IS-RRM 方案，以跟踪环境变化并相应地调整资源分配。理论分析和仿真结果都表明，我们提出的方案可以保证与事后给出的最佳策略相当的长期性能。