基于Q-learning的随机接入碰撞问题的研究  

RESEARCH ON Q-LEARNING-BASED RANDOM ACCESS COLLISION

作　　者：徐方圆 张治中[1] 李晨 Xu Fangyuan;Zhang Zhizhong;Li Chen(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《计算机应用与软件》2022年第11期119-123,140,共6页Computer Applications and Software

基　　金：重庆市市科委项目(cstc2017zdcy-zdzx0030)。

摘　　要：在5G网络中,当大量用户接入小区时会产生随机接入碰撞。对此,提出在H2H和M2M共存场景下的一种基于强化学习(Q-learning)的随机接入碰撞退避方案。研究传统的ALOHA时隙下的随机接入信道吞吐量,分析传统的ALOHA时隙的不足并介绍基于Q-learning的M2M的QL-RACH模型。在不改变协议标准的情况下,该方案既解决了M2M之间接入的问题,又不会影响H2H用户接入,减少了随机接入碰撞。将H2H分为三个接入优先等级,M2M分为两个优先等级并分析了多用户组的接入问题。仿真结果显示:优先级高的用户组在随机接入信道流量相同的情况下吞吐量高;采用Q-learning算法的吞吐量最终会收敛于一个最优值,而采用传统的SA-RACH方案的吞吐量会随着流量的增大而趋近于0。In 5G networks,random access collisions occur when a large number of users access the cell.Therefore,a random access collision avoidance scheme based on reinforcement learning(Q-learning)in the scenario of H2H and M2M coexistence is proposed.The throughput of the random access channel under the traditional ALOHA time slot was studied,and then the shortcomings of the traditional ALOHA time slot were analyzed.A QL-RACH model based on Q-learning M2M was proposed.Without changing the protocol standard,this solution not only solved the problem of access between M2M,but also did not affect H2H user access,reducing random access collisions.H2H was divided into three access priority levels,and M2M was divided into two priority levels to analyse the multi-user group priority access problem.The simulation results show that the user group with high priority has high throughput when the random access channel traffic is the same.The final throughput using the Q-learning algorithm will converge to an optimal value,while the throughput using the traditional SA-RACH scheme will approach 0 as the traffic increases.

关 键 词：Q-LEARNING 碰撞 吞吐量 多用户组 5G 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]