Airborne Radar Anti-Jamming Waveform Design Based on Deep Reinforcement Learning.

Area of Science:

• Radar Systems Engineering
• Artificial Intelligence in Defense
• Signal Processing

Background:

• Airborne radars face significant jamming and clutter, degrading performance.
• Traditional anti-jamming techniques are insufficient for modern electronic warfare.
• Improved survivability requires advanced anti-jamming waveform design.

Purpose of the Study:

• To propose a novel airborne radar waveform design method using deep reinforcement learning (DRL).
• To enhance airborne radar anti-jamming performance in complex clutter and jamming environments.
• To improve battlefield survivability and target detection capabilities.

Main Methods:

• Utilized a Markov decision process (MDP) to model the radar's operating environment.
• Employed deep neural networks and the duelling double deep Q network (D3QN) algorithm for optimal waveform strategy.
• Generated time-domain signals using an iterative transformation method (ITM).

Main Results:

• The DRL-designed waveform improved signal-to-jamming plus noise ratio (SJNR) by 2.08 dB and 3.03 dB compared to RL and LFM.
• Target detection probability increased by 26.79% and 44.25% over RL and LFM.
• Demonstrated superior anti-jamming and target detection performance at 5 W transmit power.

Conclusions:

• The proposed DRL-based airborne radar waveform design effectively enhances anti-jamming capabilities.
• The method significantly improves target detection performance in complex battlefield conditions.
• This approach offers a viable solution for airborne radar survivability against modern jamming.