Adaptive Array Shape Estimation and High-Resolution Sensing for AUV-Towed Linear Array Sonar During Turns

Junxiong Wang, Xiang Pan, Lei Cheng, Jianbo JiaoRemote 2025 · DOI

TL;DR

This paper tackles the challenge of accurately estimating the location of underwater sound sources using sonar arrays towed by autonomous underwater vehicles (AUVs), especially when the AUV is turning. They introduce a novel algorithm, ASMLM, that models the array's shape as a circular arc and jointly estimates both the array's shape and the source locations, significantly improving accuracy and reducing computational load for AUVs.

What Problem Does This Paper Address?

Autonomous underwater vehicles (AUVs) are increasingly used for various underwater tasks, often employing towed linear array sonar systems to achieve a larger aperture for better sensing capabilities. However, when an AUV turns, the towed array can bend and deform, deviating from its intended straight shape. This "shape mismatch" severely degrades the sonar system's performance, making it difficult to accurately pinpoint sound sources. Furthermore, traditional sonar methods can suffer from "left-right ambiguity," making it hard to distinguish between sources on opposite sides. Existing methods for estimating array shape are either too complex for AUVs or don't perform well under significant bending. This paper aims to solve these issues by developing a method that can accurately estimate the array's shape and source locations simultaneously, even during turns, while being computationally efficient enough for AUV deployment.

Methods

The core of the proposed solution is the Adaptive Central Angle (Shape) Marginal Likelihood Maximization (ASMLM) algorithm.

Here's a breakdown of the key methodological contributions:

  • Circular Arc Array Modeling: Instead of complex, arbitrary shapes, the authors model the towed array's deformation as a simple circular arc. This simplifies the problem by representing the array's shape with a single parameter: the central angle ($\phi$).
  • Joint Estimation Framework: The ASMLM algorithm is built upon a Bayesian estimation framework. It treats the array's central angle ($\phi$) as a hyperparameter alongside other parameters like source signal variances ($\gamma$) and noise variance ($\sigma^2$). This allows for the simultaneous estimation of both the array shape and the directions of arrival (DOAs) of sound sources.
  • Marginal Likelihood Maximization: The algorithm iteratively updates these hyperparameters. It uses a strategy where the source parameters ($\gamma$) are updated first, followed by an update of the array shape parameter ($\phi$), and this process is repeated until convergence. This iterative approach helps to manage the computational complexity.
  • Low-Complexity Sparse Bayesian Learning (SBL): To address the limited computational resources of AUVs, the ASMLM algorithm leverages sparse Bayesian learning principles. This approach is known for its high-resolution capabilities, even with limited data, and the authors have tailored it for efficiency in this context.
  • Resolution of Left-Right Ambiguity: By accurately estimating the array's bent shape, the ASMLM algorithm inherently resolves the left-right ambiguity problem without requiring additional sensors or complex post-processing.

Datasets

The paper uses simulated data for initial analysis and validation, and then presents experimental results from:

  • Sea trials in the South China Sea: A specific turning experiment was conducted using a 48-element towed linear array. The dataset consists of 120 "pings" (segments of acoustic data) collected during a turning maneuver. The trajectory of the towing platform was recorded using GPS, and the water depth was approximately 1015 meters.

Key Insights and Ideas

  • Shape Matters (A Lot): The paper strongly emphasizes that the deformation of a towed array is not a minor issue but a critical factor that can cripple traditional sonar performance. Accurately accounting for this bending is paramount.
  • Simplicity is Power: Modeling the complex array shape as a simple circular arc is a clever simplification that significantly reduces computational burden while still capturing the essential deformation.
  • Joint Estimation is Key: Trying to estimate source locations without knowing the array's shape (or vice-versa) is a recipe for poor performance. The ASMLM's ability to jointly estimate both is a major breakthrough.
  • AUV-Friendly Algorithm: The development of a computationally efficient algorithm (ASMLM) that achieves high resolution and handles array bending is crucial for making advanced sonar capabilities practical for AUVs.
  • Left-Right Ambiguity Solved: The method's inherent ability to resolve left-right ambiguity as a byproduct of accurate shape estimation is a significant advantage.

Key Results

  • Improved DOA Estimation: The ASMLM algorithm consistently provides accurate direction of arrival (DOA) estimates for sound sources throughout the turning process, outperforming conventional beamforming (CBF), MUSIC, and even the state-of-the-art ABSBL method, especially when the array experiences significant bending.
  • Accurate Array Shape Estimation: The ASMLM method accurately estimates the chord height (a measure of bending) of the towed array, closely matching the ground truth obtained from attitude sensors, particularly during pronounced turns.
  • Computational Efficiency: ASMLM demonstrates a significant speedup compared to ABSBL. The average computation time per ping is 0.3995 s for ASMLM versus 6.1530 s for ABSBL, making it much more suitable for real-time AUV applications.
  • Effective Left-Right Ambiguity Suppression: The ASMLM algorithm effectively suppresses left-right ambiguity, a problem that plagues other methods, especially as array bending increases.

Limitations and Future Work

  • Circular Arc Assumption: While effective, the circular arc model is a simplification. The paper acknowledges that the actual array shape can be more complex. Future work could explore more sophisticated shape models.
  • Computational Complexity: While ASMLM is more efficient than ABSBL, it is still more computationally intensive than basic CBF or MUSIC. Further optimizations might be explored.
  • Experimental Data: The experimental validation was conducted in the South China Sea. Further testing in diverse oceanographic conditions and with different array configurations would strengthen the findings.
  • Generalization: The paper focuses on a "steady turning process." Investigating the algorithm's performance during more dynamic and erratic maneuvers could be a valuable future direction.

Summary updated

Abstract

The deformation of the array shape during the turning process of an autonomous underwater vehicle (AUV)-towed line array sonar can significantly degrade its remote sensing performance. In this paper, a method for circular arc array modeling and dynamic deformation estimation is proposed. By treating the array shape as a hyperparameter, an adaptive central angle (shape) marginal likelihood maximization (ASMLM) algorithm is derived to jointly estimate the array shape and the directions of arrival (DOAs) of sources. The high-resolution ASMLM algorithm is used to improve the DOA estimation performance, effectively suppress left–right ambiguity and significantly reduce computational complexity, making it suitable for AUV platforms with limited computational resources. Experimental results from sea trials in the South China Sea are used to validate the superior performance of the proposed method over existing methods.
BibTeX
@inproceedings{wang2025adaptive,
  author = {Junxiong Wang and Xiang Pan and Lei Cheng and Jianbo Jiao},
  title = {{Adaptive Array Shape Estimation and High-Resolution Sensing for AUV-Towed Linear Array Sonar During Turns}},
  booktitle = {Remote},
  year = {2025},
  doi = {10.3390/rs17152690},
}
Figure