Deep Reinforcement Learning Based Decision Making for Complex Jamming Waveforms

Area of Science:

• Cognitive Electronic Warfare
• Artificial Intelligence in Defense
• Intelligent Jamming Decision Making

Background:

• Intelligent communication jamming decision-making is crucial for cognitive electronic warfare.
• Complex, non-cooperative scenarios with adaptive communication parties pose challenges for traditional reinforcement learning.
• Existing methods struggle with convergence and high interaction requirements, limiting real-world applicability.

Purpose of the Study:

• To develop an advanced deep reinforcement learning algorithm for intelligent jamming.
• To address the limitations of traditional reinforcement learning in complex electronic warfare environments.
• To enhance jamming accuracy, speed, and continuity through improved decision-making.

Main Methods:

• Proposed a novel algorithm based on deep reinforcement learning and maximum-entropy principles: Soft Actor-Critic (SAC).
• Integrated an improved Wolpertinger architecture into the SAC algorithm.
• Evaluated the algorithm in diverse jamming scenarios within a complex, non-cooperative environment.

Main Results:

• The proposed SAC algorithm demonstrated excellent performance across various jamming scenarios.
• Achieved accurate, fast, and continuous jamming, outperforming traditional methods.
• Reduced the number of interactions required, making it more suitable for real-world warfare.

Conclusions:

• The enhanced SAC algorithm with a Wolpertinger architecture offers a robust solution for intelligent jamming decision-making.
• This approach significantly improves upon traditional reinforcement learning in complex cognitive electronic warfare.
• The algorithm provides a viable method for achieving effective and efficient electronic jamming.