Port Hamiltonian

Port-Hamiltonian (pH) systems offer a powerful framework for modeling complex dynamical systems by explicitly representing energy flow and storage. Current research focuses on leveraging pH structures within data-driven modeling approaches, particularly using Gaussian processes and neural networks, to learn system dynamics from data while incorporating physical constraints and uncertainty quantification. This allows for the creation of more robust and reliable models for diverse applications, ranging from robotics and multi-physics simulations to control systems design, particularly in scenarios with limited or noisy data. The resulting models are inherently more physically consistent and offer improved stability and generalization compared to purely data-driven methods.