Jamming Strategy Optimization through Dual Q-Learning Model against Adaptive Radar.

Area of Science:

• Electronic Warfare
• Artificial Intelligence
• Radar Systems Engineering

Background:

• Modern adaptive radars employ mode switching and pulse parameter agility, increasing jamming decision complexity.
• Existing jamming methods struggle with the enhanced survivability and complexity of adaptive radars.

Purpose of the Study:

• To develop an optimized jamming strategy for adaptive radars.
• To address the increased decision-making complexity posed by radar agility.

Main Methods:

• A two-level jamming decision-making framework using dual Q-learning (DQL) was developed.
• The jamming process was modeled as a finite Markov decision process.
• High-dimensional action space was decomposed into jamming mode and pulse parameters for two interacting Q-learning models.

Main Results:

• The DQL model improved the average jamming-to-signal ratio (JSR) by 4.05%.
• Convergence time was reduced by 34.94% compared to standard Q-learning.
• The method effectively learned joint radar strategies for mode switching and parameter agility.

Conclusions:

• The proposed DQL framework offers a superior method for optimizing electronic warfare against adaptive radars.
• This approach enhances jamming effectiveness and computational efficiency in complex radar environments.