Servicenavigation

Project Details

Back

Projekt Print View

The partial relaxation method in direction-of-arrival estimation: Design and Analysis (PRIDE)

Subject Area Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
Term from 2019 to 2025
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 423747006
 
Final Report Year 2024

Final Report Abstract

The PRIDE project made significant advances in the field of Direction-of-Arrival (DoA) estimation, which is crucial for applications in radar, sonar, MIMO communications, biomedical imaging, and spectral analysis. Traditional multisource estimation methods like Maximum-Likelihood estimation, Weighted Subspace Fitting, and Covariance Matching offer high estimation performance but are computationally expensive. Simpler single-source criteria, such as conventional beamforming and the MUSIC algorithm, are computationally less demanding but perform poorly in scenarios with closely spaced sources or low signal-to-noise ratios (SNRs). The recently proposed Partial Relaxation (PR) method addresses this gap by balancing the tradeoff between performance and computational complexity. The fundamental idea of the PR method is to relax parts of the multisource array manifold during the optimization while maintaining the structure of a single steering vector. This results in simple spectral search criteria. A primary goal of the PRIDE project was to provide a theoretical foundation for the excellent threshold performance of the PR approach, which had previously been demonstrated in simulations but lacked theoretical justification. Within the PRIDE project, we derived the Cramér-Rao Bound (CRB) for the PR model, i.e., the so-called PR-CRB. This bound characterizes the DoA estimation variance of the PR estimators and at the same time also provides interesting new insights on the estimation performance of existing algorithms such as the spectral MUSIC estimator, which is also characterized by the PR-CRB. In the project, we additionally explored the asymptotic performance of the PR-DML estimator using tools from Random Matrix Theory (RMT). RMT considers the asymptotic regime where the number of snapshots and antennas both tend to infinity at a constant rate. Interestingly, asymptotic results derived from RMT often also hold in the non-asymptotic regime and, in particular, in the important threshold region where the performance of the estimators breaks down. The first-order behavior of the PR-DML cost function and its fluctuations were characterized, revealing the function converges to a Gaussian random variable. These results led to a better understanding of the resolution capabilities of the estimator and led to the development of a modified PR-DML estimator that is consistent in the considered asymptotic regime. The project also focused on PR for partly calibrated arrays. Efficient algorithms for decentralized eigenvalue decomposition were developed, which is crucial for implementing PR methods in distributed sensor arrays. In the context of joint detection and estimation, the PR framework was extended to scenarios with a large number of sources. A new DoA estimator based on PR Orthogonal Least Squares WSF was proposed in which the sources are estimated sequentially. This method improved the estimation accuracy and the computational efficiency, especially in low SNR scenarios with closely spaced sources. The PR framework was further applied to automotive MIMO radar systems, enhancing joint source detection and parameter estimation. The proposed algorithm demonstrated superior estimation performance in resolving targets using real-world measurement data. The project findings were disseminated through journal papers, conference proceedings, and tutorials.

Publications

 
 

Additional Information

© 2026 DFG Disclaimer / Copyright / Privacy Policy

Textvergrößerung und Kontrastanpassung