Modelling of Underwater Vehicles using Physics-Informed Neural Networks with Control

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Amer, Abdelhakim, Felsager, David, Brodskiy, Yury, Sarabakha, Andriy
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918002391252992
author Amer, Abdelhakim
Felsager, David
Brodskiy, Yury
Sarabakha, Andriy
author_facet Amer, Abdelhakim
Felsager, David
Brodskiy, Yury
Sarabakha, Andriy
contents Physics-informed neural networks (PINNs) integrate physical laws with data-driven models to improve generalization and sample efficiency. This work introduces an open-source implementation of the Physics-Informed Neural Network with Control (PINC) framework, designed to model the dynamics of an underwater vehicle. Using initial states, control actions, and time inputs, PINC extends PINNs to enable physically consistent transitions beyond the training domain. Various PINC configurations are tested, including differing loss functions, gradient-weighting schemes, and hyperparameters. Validation on a simulated underwater vehicle demonstrates more accurate long-horizon predictions compared to a non-physics-informed baseline
format Preprint
id arxiv_https___arxiv_org_abs_2504_20019
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Modelling of Underwater Vehicles using Physics-Informed Neural Networks with Control
Amer, Abdelhakim
Felsager, David
Brodskiy, Yury
Sarabakha, Andriy
Machine Learning
Artificial Intelligence
Robotics
Physics-informed neural networks (PINNs) integrate physical laws with data-driven models to improve generalization and sample efficiency. This work introduces an open-source implementation of the Physics-Informed Neural Network with Control (PINC) framework, designed to model the dynamics of an underwater vehicle. Using initial states, control actions, and time inputs, PINC extends PINNs to enable physically consistent transitions beyond the training domain. Various PINC configurations are tested, including differing loss functions, gradient-weighting schemes, and hyperparameters. Validation on a simulated underwater vehicle demonstrates more accurate long-horizon predictions compared to a non-physics-informed baseline
title Modelling of Underwater Vehicles using Physics-Informed Neural Networks with Control
topic Machine Learning
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2504.20019