Scholarlink

In recent years, UAV communication technology has been developing rapidly, while traditional communication signal jamming means have limited effect on UAV communication in complex electromagnetic environments. With the continuous progress of deep learning (DL) technology, the application of generative adversarial networks (GANs) in communication jamming has received widespread attention and demonstrated superior results. To this end, this paper proposes a communication jamming signal generation method based on deep generative adversarial networks. The method firstly transforms the frequency hopping signal of UAV into time-frequency two-dimensional signal samples, and normalizes the samples. On this basis, a new Variational Latent generative adversarial network (VL-GAN) network is proposed by combining the current popular LSGAN network, Variable Auto-Encoder (VAE), and introducing the perceptual loss of optimized network. Simulation results show that with the optimization of the network, the generated time-frequency 2D images are more and more similar to the real samples, which achieves better interference effects and significantly improves the robustness of the network.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

UAV Frequency Hopping Jamming Signal Generation Based on Generative Adversarial Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xianfei Han,
  • Yulin Liu,
  • Yongzhao Zhang,
  • Fenghua Xu,
  • Junsheng Mu

摘要

In recent years, UAV communication technology has been developing rapidly, while traditional communication signal jamming means have limited effect on UAV communication in complex electromagnetic environments. With the continuous progress of deep learning (DL) technology, the application of generative adversarial networks (GANs) in communication jamming has received widespread attention and demonstrated superior results. To this end, this paper proposes a communication jamming signal generation method based on deep generative adversarial networks. The method firstly transforms the frequency hopping signal of UAV into time-frequency two-dimensional signal samples, and normalizes the samples. On this basis, a new Variational Latent generative adversarial network (VL-GAN) network is proposed by combining the current popular LSGAN network, Variable Auto-Encoder (VAE), and introducing the perceptual loss of optimized network. Simulation results show that with the optimization of the network, the generated time-frequency 2D images are more and more similar to the real samples, which achieves better interference effects and significantly improves the robustness of the network.