Optimization Research on Magnetic Interference Parameter Identification and Compensation for AUV Platforms

Autonomous Underwater Vehicles (AUVs) have become essential tools for underwater magnetic surveys. However, the magnetic interference from the AUV platform severely impacts measurement accuracy. To improve the accuracy of magnetic disturbance parameter identification and compensation for AUV platfor...

Full description

Saved in:
Bibliographic Details
Main Authors: Haodong Wen, Guohua Zhou, Kena Wu, Xinkai Hu, Liezheng Tang, Shuai Xia
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10992683/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Autonomous Underwater Vehicles (AUVs) have become essential tools for underwater magnetic surveys. However, the magnetic interference from the AUV platform severely impacts measurement accuracy. To improve the accuracy of magnetic disturbance parameter identification and compensation for AUV platforms, a theoretical AUV magnetic disturbance model based on the rotating ellipsoidal shell and magnetic dipole model is established. The Success History-based Adaptive Differential Evolution with Linear Population Size Reduction (L-SHADE) algorithm is employed to optimize and identify model parameters for solving the magnetic field, and Back Propagation Neural Network (BPNN) is designed and optimized through training to enable the direct estimation of the true magnetic field from the measured magnetic field. To further improve training performance, a stacking ensemble learning (STACKING) model is introduced, with L-SHADE and BPNN as base learners and Convolutional Neural Network (CNN) as the meta-learner, integrating the advantages of both algorithms for optimization. Numerical simulations demonstrate that, under a 5° attitude error, the L-SHADE algorithm achieves mean decoding accuracies of 86.42%, 81.9%, and 86.15% for the three magnetic field components after training with low-noise data. Meanwhile, the BPNN achieves prediction accuracies of 86.58%, 83.14%, and 83.09%. By applying the stacking ensemble learning method, the maximum prediction accuracy improves by up to 25.77%. Compared to individual algorithms, the proposed fusion method significantly enhances anti-interference capability and decoding accuracy. This study provides a novel and efficient solution for AUV magnetic interference compensation and offers valuable insights for addressing similar problems.
ISSN:2169-3536