面向无人机通信网络的信道全域特性空间聚类和识别

      Space clustering and identification based on full-domain channel characteristics for UAV communication networks

      • 摘要: 为提高无人机通信网络的稳定性和可靠性,提出了一种基于信道全域特性的信道子空间聚类与识别方法。首先,利用距离域、时延域、空间域和多普勒域特性对信道进行完备表征,并提出了一种信道子空间聚类方法,将全域特性相似的信道组成信道子空间,作为无人机通信场景分类的依据。然后,提出了一种基于反向传播神经网络的信道子空间识别方法,判断新的信道数据是否属于原有信道子空间的结构,并利用信道全域特性作为特征张量以提高识别精度。同时,通过计算信道与信道子空间中心的距离,消除训练数据异常值的影响,从而提高识别的鲁棒性。最后,通过入射及反弹射线法/镜像法对176个典型数字城市场景进行仿真,获得176 000个信道的全域特性和对应信道状态信息,用于验证本文提出的聚类和识别方法的准确性。仿真结果表明,本文提出的场景识别方法可以将传统场景分类方法的176个识别目标减少至20个,且信道子空间中信道状态特性的吻合度达到99%,识别方法的准确度也达到98.7%。因此,本文提出的方法可以精确识别无人机通信工作中所处的信道子空间,为无人机通信性能优化提供依据。

         

        Abstract: In order to improve the stability and reliability of unmanned aerial vehicle (UAV) communication networks, a channel subspace clustering and identification scheme based on channel full-domain characteristics is proposed in this paper. Firstly, the channels are characterized completely using distance domain, time delay domain, spatial domain and Doppler domain characteristics, and a channel subspace clustering method is proposed to form the channel subspace of channels with similar full domain characteristics as a basis for classification of UAV communication scenarios. Then, a channel subspace identification method based on back-propagation neural network is proposed, which can determine whether the new channel data belongs to the structure of the original channel subspace and channel full-domain characteristics are used as the feature tensor to improve the identification accuracy. Meanwhile, the influence of training data outliers is eliminated by calculating the distance between the channel and the center of the channel subspace, thus improving the robustness of identification method. Finally, 176 typical digital city scenarios are simulated in this paper by shooting and bounce ray/image mirror to obtain the full-domain characteristics of 176 000 channels and the corresponding channel state information, which are used to verify the accuracy of the clustering and identification method proposed in this paper. Simulation results show that the scenario identification method proposed in this paper can reduce the 176 identification targets of the traditional scenario identification method to 20, and the accuracy of the channel state characteristics in the channel subspace and identification method reaches 99% and 98.7%. Therefore, the method proposed in this paper can accurately identify the channel subspace in the UAV communication and provide a basis for UAV communication performance optimization.

         

      /

      返回文章
      返回