基于深度强化学习的D2D通信网络抗干扰资源调度方法

Anti Interference Resource Scheduling Method for D2D Communication Network Based on Deep Reinforcement Learning

针对现有学习方法在设备到设备(device-to-device,D2D)通信网络资源调度时,抗干扰能量效率和传输速率协调分配效果不理想的问题,文章提出了基于深度强化学习的D2D通信网络抗干扰资源调度方法。分析功率控制最佳策略,建立D2D通信网络抗干扰资源调度模型;采用深度学习Q网络求解模型,构造传输速率最大化的深度神经网络(deep neural networks,DNN);以能量效率作为奖惩标准反向训练DNN,实现D2D通信网络抗干扰资源的最佳调度。实验结果表明,应用该方法后传输速率达到30 bit/s,能量效率达到3.0 Mbit/s,资源调度数为5×10^(2)~6×10^(2)个,网络吞吐量稳定在41~45 kbit/s,说明该方法能够提高传输速率和能量效率,调度性能好且网络吞吐量大。

Aiming at the problem that the anti-interference energy efficiency and transmission rate coordination allocation effect of existing learning methods are not ideal in the resource scheduling of device to device (D2D) network,an anti-interference resource scheduling method of D2D communication network based on deep reinforcement learning is proposed.The optimal power control strategy is analyzed,and the anti-interference resource scheduling model of D2D communication network is established.A deep neural networks (DNN) is constructed to maximize the transmission rate by using the deep learning Q network solution model.Taking energy efficiency as the reward and punishment standard,DNN is trained in reverse to realize the optimal scheduling of anti-interference resources in D2D communication network.The experimental results show that the energy transmission rate of this method is 30 bps,and the energy transmission efficiency reaches 3.0 Mbps,Resource scheduling number reaches 5×10^(2)~6×10^(2),and the network throughput is basically stable at 41~45 kbit/s,which shows that this method can improve transmission rate and energy efficiency,have good scheduling performance and large network throughput.