基于时空特征提取的智能网络切片算法

5G网络作为新一代的移动通信网络,提供了差异多样的各类业务服务,而通过网络切片可以将5G网络的资源虚拟化并更为有效地分配给各类服务.然而,当用户的服务需求发生变化时,需要及时地进行切片资源的重新分配和管理.为了准确适配业务需求的时空变化,提升切片性能,考虑采用深度强化学习的人工智能算法对切片资源进行实时管理,提出了基于...

Intelligent Network Slicing Algorithm Based on Temporal-Spatial Feature Extraction

As a new generation of communication network,5G network supports diversified services.Using network slicing,network resource is able to be virtualized and allocated to these services more efficiently.However,when users’requirements for services change,it’s important to reallocate and manage the resource of slices promptly.Aimed at the adaption to the temporal and spatial variability of services for enhancement of slice management,deep reinforcement learning,a kind of artificial intelligence algorithm,is adopted to manage the resource of slices in real time thus the intelligent network slicing algorithm based on temporal-spatial feature extraction is proposed.The algorithm uses Graph Attention Network(GAT)and Long Short-Term Memory(LSTM)for pre-process.Besides,it takes Deep Q-Network(DQN)for decision-making.Considering a radio access network scenario that contains several network slices over multiple base stations with several users,the superiority of this algorithm compared to conventional methods is verified through extensive simulations.