Anti-jamming optimization of cognitive radar: waveform-filter joint design and performance verification based on deep reinforcement learning

Tianyi Zhang2025 · DOI

Abstract

Aiming at the complex jamming problem of main lobe deception and sidelobe suppression in modern electronic warfare environment, this paper proposes a cognitive radar anti-jamming joint framework based on deep reinforcement learning. The framework achieves anti-jamming optimization through a three-level architecture: Interference Feature Decoupling Network (IFDN) separates the target from the jamming feature by using dual-channel feature extraction and JS divergence constraint; Dual-agent deep reinforcement learning (DADRL) collaboratively optimizes the waveform parameters and filter weight vectors. The waveform agent dynamically adjusts the transmitted waveform based on the spectrum occupancy, and the filter agent optimizes interference suppression based on the interference-to-signal ratio. Three-dimensional cognitive waveform design improve s anti-jamming and target resolution by maximizing time-frequency entropy and suppressing fuzzy function sidelobes. Experimental results show that the SINR of the framework is improved by over 26dB in the scene of complex interference, the detection probability is 0.93, and the calculation delay is 18.7ms, which is superior to the traditional methods such as STAP and CNN-LSTM. Theoretical analysis proves that the dual-agent strategy converges to Nash equilibrium, and the interference suppression performance approaches Cramér-Rao lower bound. This framework provides an efficient and real-time solution for radar anti-jamming in complex electromagnetic environment.
BibTeX
@misc{zhang2025anti,
  author = {Tianyi Zhang},
  title = {{Anti-jamming optimization of cognitive radar: waveform-filter joint design and performance verification based on deep reinforcement learning}},
  year = {2025},
  doi = {10.1117/12.3090308},
}

Loading related content...

Figure