Joint Estimation Method of DOD and DOA of Bistatic Coprime Array MIMO Radar for Coherent Targets Based on Low-Rank Matrix Reconstruction

Zhiyuan You, Guoping Hu, Hao Zhou, Guimei Zheng2022 · DOI

TL;DR

This paper introduces a novel method for simultaneously estimating the Direction of Departure (DOD) and Direction of Arrival (DOA) of multiple coherent targets using a bistatic coprime array MIMO radar. By employing virtual sensor interpolation and low-rank matrix reconstruction, the method effectively expands the array aperture, leading to improved angular resolution and estimation accuracy, especially for challenging coherent signal scenarios.

What Problem Does This Paper Address?

Estimating the DOD and DOA of targets is crucial for radar systems. However, traditional methods struggle with coherent targets, where multiple targets have highly correlated signals, leading to rank deficiency in covariance matrices and inaccurate estimations. Sparse arrays, like coprime arrays, offer advantages in terms of degrees of freedom but face similar challenges with coherent signals when using standard processing techniques. This paper tackles the problem of accurately estimating both DOD and DOA for multiple coherent targets in a bistatic MIMO radar setup, where the array geometry is inherently sparse and non-uniform.

Methods

The core of the proposed method lies in a combination of techniques:

  • Virtual Sensor Interpolation: The coprime array, with its inherent "holes" (missing sensor positions), is transformed into a virtual uniform linear array. This is achieved by interpolating virtual sensors, effectively filling the gaps and creating a denser array aperture.
  • Low-Rank Matrix Reconstruction: The covariance matrix derived from the interpolated array is often rank-deficient due to coherent signals. To address this, the paper reconstructs a full-rank covariance matrix using a low-rank matrix reconstruction approach. This involves:
  • Toeplitz Matrix Reconstruction: Spatial smoothing is applied to the covariance matrix to obtain Toeplitz matrices.
  • Convex Optimization: A kernel norm minimization theory is used to establish a convex optimization problem. This recovers the "missing" or zero elements in the reconstructed Toeplitz matrices, effectively creating an ideal, full-rank covariance matrix.
  • RD-MUSIC Algorithm: The reconstructed full-rank covariance matrix is then fed into the Reduced-Dimension Multiple Signal Classification (RD-MUSIC) algorithm. This algorithm efficiently estimates the DOD and DOA, and importantly, automatically pairs them for each target without requiring additional pairing steps.

Datasets

Not discussed in the paper. The evaluation relies on simulated data.

Key Insights and Ideas

  • Virtual Interpolation for Sparse Arrays: The paper demonstrates that virtual sensor interpolation can effectively transform a sparse, non-uniform array into a virtual uniform linear array, significantly enhancing its effective aperture.
  • Low-Rank Reconstruction for Coherent Signals: The core insight is that low-rank matrix reconstruction, particularly through convex optimization, is a powerful tool to overcome the rank deficiency issues caused by coherent targets in sparse array configurations.
  • Automatic DOD-DOA Pairing: The RD-MUSIC algorithm's ability to automatically match DOD and DOA pairs is a significant advantage, simplifying the post-processing and interpretation of results.
  • Trade-off between Accuracy and SNR: While the method excels in improving estimation accuracy and resolution, its performance is shown to be sensitive to the signal-to-noise ratio (SNR), performing less optimally in low-SNR conditions.

Key Results

The paper presents simulation results demonstrating the superiority of the proposed method (referred to as LMR algorithm) over existing methods (SMS, ASDS, TRDS) in several aspects:

  • Degrees of Freedom: The LMR algorithm shows better performance in distinguishing multiple targets, as evidenced by the clear separation of spectral peaks in Figure 3.
  • Spatial Spectrum Estimation: Figure 4 illustrates that the LMR algorithm achieves a clear spatial spectral peak map that completely overlaps with the true target angles, while other methods struggle to accurately represent the target locations.
  • Angular Resolution: Figure 5 shows that the LMR algorithm can effectively distinguish between two closely spaced targets, with no overlap in spectral peaks, outperforming other methods that exhibit overlapping peaks or poorer resolution.
  • RMSE: Figure 6 demonstrates that the LMR algorithm achieves lower Root-Mean-Square Error (RMSE) compared to other algorithms, especially at higher SNRs, indicating higher estimation accuracy.

Limitations and Future Work

  • Low SNR Performance: The paper explicitly acknowledges that the method's performance is limited under low SNR conditions due to the inherent limitations of the convex optimization algorithm.
  • Computational Complexity: While the LMR algorithm offers improved performance, it comes with higher computational complexity compared to simpler algorithms, as shown in Figure 7. However, the authors suggest this increase is "a small amount of computation" for the performance gains.
  • Future Work: The paper doesn't explicitly detail future work, but the identified limitation of low SNR performance suggests that future research could focus on developing more robust low-SNR estimation techniques or hybrid approaches. Further exploration into optimizing the convex optimization step for reduced complexity could also be a direction.

Summary updated

BibTeX
@misc{you2022joint,
  author = {Zhiyuan You and Guoping Hu and Hao Zhou and Guimei Zheng},
  title = {{Joint Estimation Method of DOD and DOA of Bistatic Coprime Array MIMO Radar for Coherent Targets Based on Low-Rank Matrix Reconstruction}},
  year = {2022},
  doi = {10.3390/s22124625},
}
Figure