基于强化学习的无人机中继网络节点轨迹优化

Nodes’Trajectory Optimization of UAV Relay Networks Based on Reinforcement Learning

无人机在远程采集和传输数据时可能遇到频谱不足的问题,此时需要借助地面网络共享频谱,即部分无人机获得地面网络提供的额外频谱作为回报,另一部分无人机为地面网络提供中继服务。针对无线网络频谱共享系统中多个无人机的飞行调度问题展开了研究,提出了一种基于Q-Learning的无人机飞行调度算法。在多个无人机进行数据中继传输时,该算法结合了放大转发和解码转发的特点,采用自适应转发模式,以最大化系统吞吐量。仿真结果表明,所提的调度算法可以对两个网络的无人机数目进行合理的分配,使每个无人机能够找到各自的最优或者次优位置,从而实现较高的系统吞吐量。

UAVs(Unmanned Aerial Vehicles)may encounter the problem of insufficient spectrum when collecting and transmitting data remotely.In this case,the ground network can be used to share the spectrum,that is,some UAVs can obtain additional spectrum provided by the ground network,and in return,others provide relay services to terrestrial networks.This paper addresses the flight scheduling problem of multiple UAVs in wireless network spectrum sharing systems,and proposes a Q-Learning based UAV flight scheduling algorithm.When multiple UAVs relay data,the algorithm combines the features of amplification forwarding and decoding forwarding,and adopts an adaptive forwarding mode to maximize system throughput.Simulation results indicate that the proposed scheduling algorithm can reasonably allocate the number of UAVs in the two networks,so that each UAV can find its own optimal or sub-optimal position,thus achieving a high system throughput.