Domain Adaptation-Based Sorting Method for UAV Swarm Targets on Multi-Station Features

Area of Science:

• Electrical Engineering
• Signal Processing
• Robotics

Background:

• Synchronous Non-Orthogonal Frequency Hopping (SNOFH) presents significant challenges for target sorting due to spectrum overlap and limited data.
• Existing sorting methods fail under these complex SNOFH conditions, necessitating novel approaches.

Purpose of the Study:

• To propose a passive sorting framework for SNOFH UAV swarm signals.
• To overcome the limitations of current methods in spectrum overlap and scarce single-source points.

Main Methods:

• Developed a spatial-location-driven sorting feature system.
• Designed a kernel joint distribution adaptation module for inter-station discrepancy elimination.
• Utilized a multi-scale wavelet-based method for sub-sampling level hopping time extraction.

Main Results:

• Achieved the highest sorting accuracy of 98% on a simulated SNOFH dataset.
• Demonstrated superior performance compared to baseline methods in most SNOFH scenarios.
• Exhibited robust performance against noise, clock errors, frequency offsets, and multipath effects.

Conclusions:

• The proposed passive sorting framework is effective for SNOFH UAV swarm signals.
• The method offers high accuracy and robustness, suitable for regular and slow-varying UAV formations.
• Reduced reliance on prior frequency hopping parameters and hardware fingerprints.