Abstract
In radar accurate localization of unmanned aerial vehicle (UAV) swarms, the high density, similar motion parameters, small radar cross-section (RCS), strong noise, and far range put forward high requirements on radar resolution and transmitting power. This paper presents a super-resolution framework for radar accurate localization of UAV swarms without changing radar hardware. By combining long-time integration (LTI) with a gridless sparse method, specifically Frequency-Selective Reweighted Atomic-Norm Minimization (FSRAM), the proposed method achieves superior range resolution and noise robustness. Simulations and real-world X-band radar experiments demonstrate that this approach significantly outperforms traditional methods like MUSIC and standard RAM in resolving dense swarms under noisy conditions.
Introduction: The Challenge of UAV Swarm Surveillance
Unmanned Aerial Vehicle (UAV) swarms present a significant challenge for traditional surveillance radar systems. Drones typically have a small Radar Cross-Section (RCS), reflecting minimal energy. When flying in dense formations with similar velocities, they often fall within the same "resolution cell" of a standard radar.
Standard radar systems are physically limited by their bandwidth (determining range resolution) and observation time (determining Doppler resolution). If multiple drones are closer than these physical limits, they appear as a single, indistinguishable target. While hardware upgrades like increased bandwidth could solve this, they are often cost-prohibitive. This research proposes a signal-processing solution to achieve super-resolution using existing hardware.
The Three-Step Super-Resolution Framework
The core of the proposed solution is a three-step framework that integrates Long-Time Integration (LTI) with gridless sparse recovery techniques. This fusion allows the system to "break" traditional resolution limits by treating target localization as a continuous spectral recovery problem.
- Step 1: Detection and Coarse Estimation: The system applies the Keystone Transform (KT) to compensate for range cell migration, where a target moves across range bins during the observation period. This step integrates energy over time to detect the swarm and provide initial estimates of position and velocity.
- Step 2: SNR Enhancement and Doppler Localization: By "staring" at the detected swarm for an extended duration, the radar boosts the Signal-to-Noise Ratio (SNR). This enhanced data allows for better localization along the Doppler dimension, separating drones with even minor differences in radial velocity.
- Step 3: Range Super-Resolution via FSRAM: For drones still clumped in the same range/Doppler bins, the authors introduce Frequency-Selective Reweighted Atomic-Norm Minimization (FSRAM). Unlike "on-grid" methods that suffer from grid mismatch, this "gridless" approach treats locations as continuous variables. The "frequency-selective" component uses the coarse data from Step 2 as prior knowledge, narrowing the search space and increasing robustness against noise.
Experimental Validation and Results
The methodology was validated through both extensive numerical simulations and real-world experiments.
Numerical Simulations
Simulations tested the algorithm against varying SNRs and target densities. The FSRAM algorithm maintained a high success rate in separating adjacent UAVs even at an SNR as low as -13dB. In comparison, traditional subspace methods like MUSIC and standard reweighted atomic-norm minimization (RAM) failed to resolve targets under similar noise levels or when signals were highly correlated.
Real-World X-Band Radar Experiment
To verify practical effectiveness, the researchers conducted an experiment using an X-band radar with 8 antenna elements and a 10MHz bandwidth. The system tracked three actual UAVs flying in close proximity.
As shown in the experimental results, the proposed method successfully resolved drones that were flying in the same range cell and at nearly identical velocities—scenarios where traditional detection would have identified only a single target peak.
Conclusion and Future Directions
The proposed range super-resolution framework offers a robust, software-driven path to enhancing radar capabilities for UAV swarm tracking. By leveraging LTI to provide prior knowledge for gridless sparse recovery, the system achieves high accuracy without hardware modifications.
Future work will focus on addressing the computational complexity of solving Semi-Definite Programming (SDP) problems, potentially using faster implementation strategies like the Alternating Direction Method of Multipliers (ADMM). Additionally, the researchers aim to extend the model to account for maneuvering targets that do not maintain constant velocity.
Summary updated