[Paper Review] Angular Resolution of Closely-Spaced Targets with Antenna Arrays

Ulrich Nickel, David Crouse|arXiv (Cornell University)|Aug 14, 2019

Advanced SAR Imaging Techniques17 references4 citations

TL;DR

This paper presents a novel method for resolving closely-spaced targets using antenna arrays by formulating angular resolution as a statistical estimation and detection problem. It introduces the Q-function minimization framework and stochastic approximation techniques to achieve superresolution, demonstrating theoretical convergence and robustness under realistic array impairments such as noise, coupling, and quantization.

ABSTRACT

This is an English translation of Ulrich Nickel's PhD dissertation with the original title "Winkelauflösung eng benachbarter Ziele mit Gruppenantennen." It describes maximum-likelihood angular superresolution of closely-spaced targets. It also discusses estimating the number of targets present.

Motivation & Objective

• To address the fundamental challenge of resolving closely-spaced targets beyond the classical beamwidth limit using array signal processing.
• To develop a statistical framework that treats angular resolution as a detection and estimation problem under uncertainty.
• To provide a mathematically rigorous solution for target resolution using minimum variance distortionless response (MVDR) and information-theoretic principles.
• To analyze the robustness of the resolution method under realistic array impairments such as coupling, correlated noise, and quantization effects.
• To establish theoretical convergence and performance bounds for the proposed stochastic approximation algorithm in planar arrays.

Proposed method

• Formulates the angular resolution problem as minimizing a Q-function derived from the MVDR beamformer response.
• Applies stochastic approximation to iteratively minimize the Q-function using spatial samples, ensuring convergence under mild conditions.
• Introduces a multihypothesis test based on the averaged Q-statistic to detect the presence of multiple targets.
• Uses Hermite forms and characteristic functions to model the probability distribution of the Q-statistic under Gaussian noise.
• Implements grid search with averaging to improve resolution and reduce false alarms in low SNR conditions.
• Analyzes perturbations from coupling, extended targets, correlated interference, and amplifier fluctuations to validate robustness.

Experimental results

Research questions

• RQ1 Can superresolution be achieved in antenna arrays using only spatial samples without prior knowledge of target parameters?
• RQ2 What is the theoretical convergence behavior of the stochastic approximation algorithm for Q-function minimization in angular resolution?
• RQ3 How does the proposed multihypothesis test compare to classical likelihood ratio tests in detecting closely-spaced targets?
• RQ4 To what extent are resolution performance and estimation accuracy degraded by array imperfections such as coupling and correlated noise?
• RQ5 What is the impact of signal bandwidth and target extension on the achievable angular resolution?

Key findings

• The Q-function has a unique minimum at the true direction of arrival (DOA) under ideal conditions, ensuring consistent estimation.
• Stochastic approximation converges almost surely to the global minimum of the Q-function, enabling robust DOA estimation.
• The multihypothesis test based on the ¯Q statistic achieves detection performance close to the theoretical optimum under Gaussian noise.
• The method remains stable and accurate under correlated noise, amplifier fluctuations, and quantization, with bounded error growth.
• Theoretical analysis confirms that the Q-function's variance is trace(ΓRΓR), providing a measure of estimation uncertainty.
• Numerical results show that the averaged grid search improves resolution by suppressing noise-induced fluctuations and reducing false alarms.