基于深度强化学习的工业物联网多用户频谱分配

随着5G时代的来临,工业物联网将迎来蓬勃发展.然而,联网设备数量的不断增加,加剧了有限的频谱资源与大量的通信需求之间的矛盾.针对以上问题,提出了一种基于聚类分组和深度强化学习的合作式动态频谱分配算法,使用户可以获得较低的信息传输中断概率以及较少的多跳转发次数,快速找到信息传输的最优路径.在动态频谱分配中,该算法可以有效降低主、次用户信道接入的碰撞概率,提升频谱资源的利用率.对于少部分计算能力有限的用户,通过协调同组次用户的计算能力来完成策略的训练,实现了计算资源的高效利用.经过多次仿真实验验证,所提出的联合算法与现有的方法相比具有更高的信道利用率和更低的用户接入碰撞率.

Multi-user spectrum allocation using deep reinforcement learning for industrial IoT

With the advent of the 5G era,the technology of industrial Internet of Things(IIoT) is springing up vigorously.However,as the number of Internet devices is constantly rising,conflict between limited spectrum resources and high demand for communication is intensified.For the above problems,this paper proposes a dynamic grouping strategy based on a new grouping clustering mode and deep reinforcement learning(DRL) algorithm.The algorithm can relatively lower interruptive possibility of information transmission and multiple jumps for users who can efficiently find the optimal path of information transmission.Furthermore,in the process of dynamic spectrum allocation,the algorithm can effectively reduce the collision probability and promote the utilization of primary and secondary user of spectrum resources channel.Additionally,for a small portion of those users,with limited computing ability,they can complete the strategy training by cooperating with other users of the same groups,thus realizing the efficient utilization of computing resources.Numerous simulation experiments suggest that,compared with methods on hand,the proposed joint algorithm has the advantages over higher rate channel utilization and lower collision probability of user access.