基于强化学习的异构网络流量卸载方案研究

Research on traffic unloading scheme of heterogeneous network based on reinforcement learning

面对未来爆炸性的数据流量增长问题,提出了一种基于强化学习思想的流量卸载的方案。在此基础上,对算法进行改进,提出一种基于随机竞争的算法。通过搭建异构网络下的MATLAB仿真平台,对LTE系统,强化学习切换算法的系统和随机竞争切换算法的系统进行仿真对比分析。结果表明,采用强化学习算法和随机竞争算法的系统的性能比较优秀,总体吞吐量提升2倍左右;此外,对比强化学习算法与随机竞争算法的仿真结果,总体上改进后的随机竞争算法的性能比强化学习算法的性能优秀,但在小区与用户数量较少的情况下,强化学习算法的系统网络性能最佳。

In order to face the future explosive growth of user traffic data problems,this paper proposed a scheme of traffic offloading based on reinforcement learning. Than the algorithm of reinforcement learning is improved and an algorithm of random competition is proposed. Through MATLAB simulation platform based on heterogeneous networks,we make a careful comparative analysis with the LTE system and the reinforcement learning algorithm and the random competition algorithm. The results show that： For the heterogeneous network using the reinforcement learning algorithm and the random competition algorithm,the throughput of the network is about twice that of the LTE network. In addition,by comparing the reinforcement learning algorithm with the random competition algorithm simulation results,it can be seen that the heterogeneous network using random competition algorithm has improved signal quality and throughput performance. However,in the case of small number of cells and users,the performance of the reinforcement learning algorithm is the best on the heterogeneous network.