TL;DR
This paper tackles the challenge of accurately locating targets with Frequency Diverse Array (FDA)-MIMO radar in the presence of both traditional Gaussian jamming (barrage) and more realistic, non-Gaussian impulsive jamming (burst). They introduce a novel "low-rank + low-rank + sparse" decomposition model and a two-step GoDec algorithm to effectively separate the desired target signal from these mixed interferences, leading to more robust and accurate localization.
What Problem Does This Paper Address?
Radar systems, particularly FDA-MIMO radars, are crucial for target localization. However, their performance can be severely degraded by jamming signals. While existing methods often assume jamming is Gaussian (barrage jamming), real-world scenarios frequently involve non-Gaussian impulsive or burst jamming, which can arise from various sources like amplifier saturation, sensor failures, or even friendly equipment. This non-Gaussian nature, characterized by heavier tails in its probability distribution, can significantly impair the accuracy of estimators relying on Gaussian assumptions. This paper addresses the critical need for a radar signal processing technique that can effectively handle and remove both barrage and burst jamming simultaneously for accurate target localization.
Methods
The core of the paper's contribution lies in its novel signal decomposition model and a corresponding algorithm.
- "Low-rank + Low-rank + Sparse" Decomposition Model: Instead of traditional "low-rank + sparse" models, this paper proposes a more comprehensive model that explicitly accounts for three components:
- A low-rank component representing the desired target signal.
- Another low-rank component for barrage jamming.
-
A sparse component for burst jamming. This model is a more accurate representation of the complex interference environment.
-
Two-Step GoDec Algorithm: To solve the decomposition problem, the authors adapt the existing "Go Decomposition" (GoDec) method. Their proposed two-step approach uses alternating minimization and leverages a priori knowledge of the rank of the target signal and barrage jamming components. This strategy is designed to suppress false targets that might arise from the mixed jamming and is parameter-free, making it easier to implement. The algorithm iteratively refines estimates of the low-rank target signal, low-rank barrage jamming, and sparse burst jamming components.
Datasets
The paper does not introduce new datasets. Instead, it relies on simulated radar data generated according to the described signal and jamming models. The parameters used for simulation are detailed in Table I, which specifies characteristics like reference frequency, frequency increment, bandwidth, interference angle, INR, number of elements, element spacing, and number of pulses.
Key Insights and Ideas
- The Reality of Mixed Jamming: The paper highlights that assuming only Gaussian jamming is an oversimplification. The inclusion of non-Gaussian burst jamming is crucial for practical radar systems.
- "Low-rank + Low-rank + Sparse" is Key: The proposed decomposition model is a significant step forward because it explicitly separates the target signal from two distinct types of interference (barrage and burst), each with different statistical properties.
- Two-Step GoDec for Robustness: The two-step GoDec algorithm, by incorporating a priori rank information, effectively distinguishes between the desired signal and the various jamming components, leading to improved robustness against mixed interference.
- FDA's Advantage: The paper reiterates the benefit of FDA-MIMO radar in separating targets with the same angle but different ranges, a capability that is preserved and leveraged in their proposed method.
Key Results
- Improved Localization Accuracy: Simulation results demonstrate that the proposed method can accurately estimate target range and angle even in the presence of significant barrage jamming. Figures 8 and 9 show that the Root Mean Square Error (RMSE) for range and angle estimation is comparable to the Cramér-Rao Bound (CRB) under various jamming conditions.
- Robustness to Jamming Direction: The method shows good performance when the jamming angle is different from the target angle. The probability of success in angle and range estimation (Figures 10 and 11) is high (approaching 100%) for moderate SNRs when the jamming angle is sufficiently far from the target angle.
- Separation of Targets: Figure 5 illustrates the successful separation of two targets with different locations. Figures 6 and 7 demonstrate the ability to distinguish targets with the same range but different angles (FDA's strength) and targets with the same angle but different ranges (MIMO's strength).
- Convergence: The proposed two-step GoDec method is shown to converge quickly, often within a single iteration in simulations (Figure 4).
Limitations and Future Work
- Performance Degradation with Co-located Jamming: The paper acknowledges that the proposed method struggles to estimate target parameters when the barrage jamming angle is very close to or the same as the target angle. This is a significant limitation, and the authors suggest it as a direction for future work.
- A Priori Rank Information: While the algorithm is parameter-free in terms of iterative optimization, it relies on a priori knowledge of the rank of the target signal and barrage jamming components. Determining these ranks accurately in real-world scenarios can be challenging.
- Future Work: The authors explicitly mention the need for an unbiased joint range and angle estimator that can handle cases where the target and jamming angles are coincident. They also suggest exploring how to improve parameter estimation accuracy, particularly in challenging scenarios.
Summary updated