Balancing accuracy and convergence rate: a hybrid optimisation algorithm for parameter identification of unmanned marine vehicles
In this paper, a novel and well-balanced hybrid optimisation algorithm called NMA is proposed for parameter identification of a coupled three-degree-of-freedom unmanned marine vehicle in the presence of measurement noise. Firstly, a multi-step iterative prediction model is designed as identification...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Engineering Applications of Computational Fluid Mechanics |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/19942060.2025.2503779 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In this paper, a novel and well-balanced hybrid optimisation algorithm called NMA is proposed for parameter identification of a coupled three-degree-of-freedom unmanned marine vehicle in the presence of measurement noise. Firstly, a multi-step iterative prediction model is designed as identification structure, which uses the previous prediction as the current input instead of the measurement, thereby ensuring that the predictions are not affected by measurement noise. In this structure, the loss function incorporates residuals from all degrees of freedom, thus enabling a comprehensive optimisation of the parameters. Secondly, the proposed NMA algorithm integrates the Adam algorithm to guide the search of the Nelder-Mead (NM) simplex algorithm. This integration not only enhances the convergence rate but also improves the accuracy, ultimately achieving satisfactory optimisation performance. Thirdly, zero-order gradient estimators are introduced to replace the finite difference method in the original NMA algorithm, reducing the computational cost and resulting in a variant called ZO-NMA. Finally, the simulation results demonstrate that the proposed NMA algorithm significantly outperforms the individual algorithms. Specifically, it reduces the computation time by 27.09% compared to the existing Adam algorithm, while improving the accuracy by 13.2 percentage points compared to the standard NM simplex algorithm. Moreover, ZO-NMA further reduces computation time by 19.05% compared to the NMA algorithm. |
|---|---|
| ISSN: | 1994-2060 1997-003X |