%0 Journal Article %T 基于SSAE-LSTM的无人机时–空域频谱预测研究
Research on Spatio-Temporal Spectrum Prediction for UAVs Based on SSAE-LSTM %A 李鹏达 %A 陈佳美 %A 周晓达 %A 王宇鹏 %J Hans Journal of Wireless Communications %P 45-58 %@ 2163-3991 %D 2025 %I Hans Publishing %R 10.12677/hjwc.2025.153006 %X 无人机等用频设备的激增使得本就有限的频谱资源变得更加稀缺。采用认知无线电技术，认知无人机如果可以在频谱感知之前进行准确的频谱预测，则可以很大程度提高频谱利用率。现有的针对固定节点的预测方法不能反应高度动态的无人机频谱预测的空间变化属性。本文提出一种基于堆叠稀疏自编码器(SSAE)与长短期记忆网络(LSTM)的无人机时–空域频谱预测方法(SSAE-LSTM)，使用真实测量的频谱数据集进行频谱预测。首先，建立时–空域系统模型，将无人机历史实际测量的频谱数据作为底层数据输入SSAE逐层训练，获取频谱隐藏特征表达。然后，将提取的特征序列输入到LSTM网络中进行无人机频谱的长期预测。实验结果表明，所提出的方法能够有效预测认知无人机通信环境中的频谱使用情况，在预测精度和预测误差方面优于现有的无人机频谱预测方法。
With the proliferation of frequency-using devices such as unmanned aerial vehicles (UAVs), the already limited spectrum resources have become even scarcer. By employing cognitive radio technology, cognitive UAVs can significantly improve spectrum utilization if they can accurately predict the spectrum before spectrum sensing. Existing prediction methods for fixed nodes fail to capture the spatial variation attributes of highly dynamic UAV spectrum prediction. This paper proposes a temporal-spatial spectrum prediction method for UAVs based on Stacked Sparse Auto-encoder (SSAE) and Long Short-Term Memory (LSTM) networks (SSAE-LSTM), using real-measured spectrum datasets for spectrum prediction. Firstly, a temporal-spatial system model is established, the historical and actual spectrum data measured by UAVs is input into the SSAE for layer-by-layer training to obtain the hidden feature representation of the spectrum. Then, the extracted feature sequences are input into the LSTM network for long-term spectrum prediction for UAVs. Experimental results demonstrate that the proposed method can effectively predict the spectrum usage in cognitive UAV communication environments, outperforming existing UAV spectrum prediction methods in terms of prediction accuracy and prediction error. %K 无人机频谱预测， %K 时– %K 空域频谱预测， %K 堆叠稀疏自编码器， %K 长短期记忆网络
UAV Spectrum Prediction %K Temporal-Spatial Spectrum Prediction %K Stacked Sparse Auto-Encoder %K Long Short-Term Memory Network %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=116943