TL;DR
This paper introduces a novel sparse learning framework for detecting multiple noise-like jammers (NLJs) in radar systems. By leveraging the inherent sparsity of NLJ signals, the proposed method achieves robust detection and accurate estimation of jammer parameters, outperforming existing techniques.
What Problem Does This Paper Address?
Modern radar systems are increasingly vulnerable to electronic warfare, particularly from noise-like jammers (NLJs). These jammers blend into the radar's thermal noise, degrading its sensitivity and making it difficult to detect targets or even identify that jamming is occurring. While traditional methods like Sidelobe Blankers (SLB) are effective against coherent jammers, they are ineffective against NLJs. Existing Electronic Counter-Countermeasure (ECCM) techniques, such as Sidelobe Cancelers (SLC) and Adaptive Digital BeamForming (ADBF), can mitigate NLJs but often come with significant computational burdens and may require additional hardware. Furthermore, accurately detecting multiple NLJs and estimating their precise angles of arrival (AoAs) simultaneously is a challenging problem, as classical maximum likelihood approaches become computationally intractable. This paper aims to address this gap by developing an efficient and accurate method for joint detection and parameter estimation of multiple NLJs.
Methods
The authors propose a systematic framework based on a sparse representation of the problem. The core idea is to exploit the fact that while the number of jammers ($N_j$) is typically small compared to the number of potential angular locations ($L$) in the surveillance sector. This leads to a sparse signal model.
Key methodological components include:
- Likelihood Ratio Test (LRT): The detection decision is based on the LRT, comparing the probability of observing the data under the jamming hypothesis ($H_1$) versus the noise-only hypothesis ($H_0$).
- Cyclic Optimization: To handle the computational complexity of estimating multiple jammer parameters, the authors employ cyclic optimization procedures. These iterative algorithms refine estimates of the jammer parameters (specifically, a sparse vector representing their presence and power at different angles) and the noise variance.
- Sparsity-Promoting Prior: A prior distribution is introduced that encourages sparsity in the jammer parameter estimates, reflecting the assumption that only a few angles are actually occupied by jammers. This prior is crucial for guiding the optimization process.
- Two Architectures:
- SC-LRT (Sparse Cyclic LRT): This architecture assumes the thermal noise power ($\sigma_n^2$) is known and estimates the jammer parameters.
- SDC-LRT (Sparse Doubly Cyclic LRT): This more advanced architecture jointly estimates both the jammer parameters and the thermal noise power, offering greater adaptivity at the cost of increased computational load.
- Model Order Selection (MOS): Techniques like the Bayesian Information Criterion (BIC) are used to estimate the number of jammers from the sparse parameter estimates.
Datasets
The paper evaluates the proposed methods using simulated data. No specific datasets are mentioned as being publicly released or used for training. The authors generate synthetic data to mimic realistic radar scenarios with varying numbers of NLJs, signal-to-noise ratios (SNR), and angular sampling resolutions.
Key Insights and Ideas
- Sparsity is Key: The fundamental insight is that the problem of detecting multiple NLJs can be effectively framed as a sparse recovery problem. By treating the angular sector as a discrete grid and assuming only a few locations are occupied by jammers, the problem becomes tractable.
- Joint Detection and Estimation: The proposed framework elegantly combines the detection of NLJs with the estimation of their parameters (AoA and power) in a single, unified approach, avoiding separate stages that can introduce errors.
- Suboptimal but Practical: The authors acknowledge that their approach is suboptimum compared to a hypothetical ideal Maximum Likelihood approach (which is intractable). However, the suboptimum nature is a necessary trade-off for achieving practical computational feasibility.
- Adaptive Noise Estimation: The SDC-LRT's ability to adaptively estimate the thermal noise power ($\sigma_n^2$) makes it more robust in scenarios where the noise level might fluctuate.
- Improved Angular Resolution: The sparsity-based estimation is shown to potentially improve angular resolution, especially for high NLJ powers.
Key Results
The proposed SC-LRT and SDC-LRT algorithms demonstrate superior performance compared to a state-of-the-art competitor (SPICE-LRT) in various scenarios:
- Detection Performance (Pjd): The proposed methods achieve better Probability of Jammer Detection (Pjd) for a given Probability of False Jammer Detection (Pfjd), often with a gain of around 2 dB.
- Parameter Estimation: The algorithms provide accurate estimates of the number of jammers and their AoAs. The RMS error for the Hausdorff distance between true and estimated jammer parameters is consistently lower for the proposed methods.
- Classification Accuracy: The proposed methods show a higher percentage of correct jammer count estimation, especially in challenging scenarios (e.g., finer grid sampling), while the SPICE-LRT tends to overestimate the number of jammers.
- Robustness to Grid Sampling: Performance generally improves with wider sampling intervals, as it reduces ambiguity and energy spillover between grid points.
- Effect of NLJ Number: The performance is maintained even when the number of NLJs increases from 3 to 4.
Limitations and Future Work
- Suboptimal Approach: The authors explicitly state that their methodology leads to suboptimum solutions due to the mathematical intractability of the optimal Maximum Likelihood Approach.
- Computational Complexity: While more efficient than MLA, the iterative nature of the proposed algorithms still incurs a computational burden, especially for the SDC-LRT.
- Fusion Strategies: The paper notes that artifacts like "ghosts" can arise from energy spillover. Future work will focus on designing enhanced fusion strategies to mitigate these artifacts and improve grid resolution.
- Extension to Other Scenarios: The authors suggest extending the framework to handle multiple coherent targets and to new contexts like 5G systems utilizing phased arrays.
Summary updated