Reinforcement Learning Based Resource Allocation for Network Slicing in 5G C-RAN

In network slicing enabled 5G cloud radio access networks (C-RAN), we study the network slice resource allocation to ensure the differentiated performance requirements of diversified services from mobile networks and improve the revenue of operators. Considering the characteristics of 5G C-RAN architecture, we propose a network slice resource allocation framework, which is composed of an upper layer, which performs the mapping of virtual protocol stack functions; and a lower layer, which manages radio remote unit (RRU) association, subchannel and power allocation. We model a utility maximization problem based on the proposed framework. Then we proposed a reinforcement learning based two-stage network slice resource allocation algorithm, which uses the multi-agent Q-learning process to reduce the complexity of the Q-value table. Simulation results demonstrate that the proposed algorithm can improve the whole network utility while ensuring the network performance of the virtual network operators compared with the baseline schemes.