基于增强学习的5G网络切片资源动态优化方案

[摘要] 提出了一种基于增强学习的网络切片资源动态优化方案。使用该方案动态调整网络切片资源时，通过考虑未来网络切片中的业务流量变化情况，对业务流量进行预测，从而推断出未来网络资源的划分情况；再通过增强学习算法，使得未来时刻的网络资源划分状态对当前划分策略做出影响，从而得到当前的最佳策略。基于该算法，可以保证在资源分配过程中对网络需求变化做出快速响应，并通过仿真进行了验证。

[关键词] 5G；网络切片；增强学习；动态优化

[Abstract] In this paper, a dynamic optimization algorithm based on reinforcement learning for network slicing division is proposed. Network resources can be dynamically allocated in the following ways: the traffic flow can be predicted by considering the changes of flow, then the division of future network resources can be deduced; based on reinforcement learning algorithm, the current partition strategy will be affected by the state of network resource partitioning in the future, and the best division strategy can be got. Based on this algorithm, the change of network requirements can be rapidly responded in the process of resource allocation, and verified by simulation.

[Keywords] 5G; network slicing; reinforcement learning; dynamic optimization