An Autonomous Transmission Scheme Using Dueling DQN for D2D Communication Networks
An Autonomous Transmission Scheme Using Dueling DQN for D2D Communication Networks
复制标题
DOI：
10.1109/tvt.2020.3041458
复制
发表时间：
2020-12
影响因子：
6.8
通讯作者：
T. Ban
中科院分区：
文献类型：
--
作者：
T. Ban
在本文中，我们研究设备到设备（D2D）通信网络，这是下一代移动的通信网络和许多其他应用，如无人机（UAV），车对车（V2V）和物联网（IoT）的关键技术之一。在我们的研究中考虑的覆盖D2D通信网络使用与蜂窝网络使用的无线电资源分离的专用无线电资源，并且在D2D网络中存在共信道干扰，而在两个网络之间没有交叉信道干扰。我们提出了一种用于覆盖D2D网络的新传输方案，该方案使用决斗深度强化学习（DRL）架构。DRL在动作不影响后续状态的环境中特别有效，如在无线通信网络中。本文的主要贡献是，所提出的架构被设计为仅利用每个D2D设备可以通过测量信道容易地获得的信息。因此，所提出的方案使得D2D设备能够训练它们自己的神经网络，并自主决定是否传输数据，而无需来自基础设施的任何干预。所提出的方案的性能进行了分析，在平均和率，并比较三个基线计划。仿真结果表明，该方案可以实现几乎最佳的和速率与低信噪比（SNR）值，而无需任何干预的基础设施。
In this paper, we investigate device-to-device (D2D) communication networks which are one of the key technologies for next-generation mobile communication networks and many other applications such as unmanned aerial vehicles (UAVs), vehicle-to-vehicle (V2V), and Internet of things (IoT). The overlay D2D communication networks that are considered in our study use dedicated radio resources separate from what cellular networks use and there exists co-channel interference in D2D networks without cross-channel interference between two networks. We propose a new transmission scheme for overlay D2D networks that uses a dueling deep reinforcement learning (DRL) architecture. The DRL is especially effective in environments where actions do not affect subsequent states as in wireless communication networks. The main contribution of this paper is that the proposed architecture is designed to utilize only information that each D2D devices can easily obtain by measuring channels. The proposed scheme thus enables D2D devices to train their own neural networks and to decide autonomously whether to transmit data without any intervention from infrastructures. The performance of the proposed scheme is analyzed in terms of average sum-rates and is compared to three baseline schemes. Simulation results show that the proposed scheme can achieve almost optimal sum-rates with low signal-to-noise (SNR) values without any intervention from infrastructure.