Parameter identification of vertical plane model for autonomous underwater vehicle based on hierarchical multi-innovation stochastic gradient algorithm

• Published in: Measurement : journal of the International Measurement Confederation Vol. 252; p. 117316
• Main Authors: Liu, Yang, Wang, Longjin, An, Shun, Liu, Peng, Fan, Zhimin
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2025
• Subjects: Autonomous underwater vehicle; Hierarchical principle; Multi-innovation stochastic gradient; Parameter identification; Vertical plane model; Hierarchical principle; Multi-innovation stochastic gradient; Parameter identification; Vertical plane model; Autonomous underwater vehicle
• ISSN: 0263-2241
• DOI: 10.1016/j.measurement.2025.117316

In this paper, the problem of online parameter identification of the vertical plane motion model for autonomous underwater vehicle (AUV) is investigated. The AUV model is processed based on the Euler discretization principle, and an AUV discretization model suitable for parameter identification is obtained. Aiming at the shortcomings of the traditional stochastic gradient (SG) algorithm in terms of accuracy, a multi-innovation stochastic gradient (MI-SG) algorithm is introduced to improve the accuracy of AUV parameter identification. To further improve the identification efficiency and accuracy, the AUV model is decomposed into two sub-models of smaller dimensions through the principle of hierarchical identification, and a hierarchical multi-innovation stochastic gradient (H-MI-SG) algorithm is developed. The H-MI-SG algorithm not only inherits the high accuracy features of MI-SG algorithm, but also effectively improves the convergence speed and accuracy of the identification algorithm through the hierarchical processing. Simulation results show that the H-MI-SG algorithm has better performance with 50% reduction in convergence time compared to MI-SG algorithm and 6.3% improvement in convergence accuracy compared to MI-SG algorithm. •Based on the Euler discretization idea, the AUV vertical plane discretization model is derived.•To improve the convergence speed, a hierarchical multi-innovation stochastic gradient algorithm is proposed.•The proposed system identification algorithm has online identification capability.