Modelling of Underwater Vehicles using Physics-Informed Neural Networks with Control
This work addresses modeling challenges for underwater vehicles, but it appears incremental as it extends existing PINNs with control integration.
The authors tackled the problem of modeling underwater vehicle dynamics by introducing the Physics-Informed Neural Network with Control (PINC) framework, which improved long-horizon prediction accuracy compared to a non-physics-informed baseline.
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