Title | ||
---|---|---|
Deep reinforcement learning-based radio function deployment for secure and resource-efficient NG-RAN slicing |
Abstract | ||
---|---|---|
Network functions virtualization is a prominent technology for next-generation radio access network (NG-RAN) slicing to achieve customization for various vertical services, such as auto-manufacturing and auto-driving. However, when virtualized radio function blocks with different security levels of services share a common server to achieve network resource-saving, a co-resident threat is exposed due to the lack of physical isolation. Therefore, designing a secure and efficient NG-RAN slicing strategy is necessary but difficult, especially for such distributed networks in which different tenants have distinct network bandwidth and security level requirements. To this end, we first formulate this slicing problem as an ILP model called Secure isolation and resource Efficiency-oriented Multi-Objective NG-RAN Slicing (SEMONS). However, the SEMONS is not practical for online execution due to its high computational complexity. Then, we propose a secure and efficient RAN slicing method for fast online execution based on a deep reinforcement learning (DRL) framework. The Wolpertinger policy algorithm is leveraged to train the agent in the DRL framework, which combines the deep deterministic policy gradient with the K-Nearest-Neighbor algorithm to reduce the size of the exploration space. Then the training complexity is reduced, and the learning results are optimized. Extensive simulation results show that the DRL framework can obtain the near-optimal secure and efficient NG-RAN slicing strategies compared to the SEMONS model, with only 6% target deviation on average, and it also outperforms the greedy baseline algorithm by 14.5%. |
Year | DOI | Venue |
---|---|---|
2021 | 10.1016/j.engappai.2021.104490 | Engineering Applications of Artificial Intelligence |
Keywords | DocType | Volume |
Deep reinforcement learning,Wolpertinger policy,NG-RAN slicing,Security,Resource efficiency | Journal | 106 |
ISSN | Citations | PageRank |
0952-1976 | 0 | 0.34 |
References | Authors | |
0 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Pengfei Zhu | 1 | 249 | 31.05 |
Jiawei Zhang | 2 | 1 | 0.69 |
Yuming Xiao | 3 | 2 | 1.73 |
Jiabin Cui | 4 | 0 | 0.34 |
Lin Bai | 5 | 0 | 0.68 |
Yuefeng Ji | 6 | 13 | 8.67 |