Search papers, labs, and topics across Lattice.
This paper introduces a Bayesian filtering approach to learn Lagrangian dynamics from noisy, partial measurements by modeling system dynamics with a Lagrangian mechanics framework. The method parameterizes kinetic and potential energies using neural networks and treats unknown external forces as white Gaussian noise, leading to a stochastic state-space model that captures the dynamics. Experimental results on pendulum and Duffing oscillator examples show that this approach outperforms conventional Lagrangian neural networks and approximate Bayesian filters, highlighting its robustness in handling noise in measurements.
Learning Lagrangian dynamics from noisy measurements can be achieved more effectively with a Bayesian filtering approach than traditional methods.
This paper proposes a Bayesian filtering-based approach for learning the dynamics of a physical system from partial, noisy measurements. We model the system dynamics using a Lagrangian mechanics formulation. As in Lagrangian neural networks (LNNs), we parameterize the kinetic and potential energies with neural networks. The unknown external forces in the Lagrangian formulation are modeled as white Gaussian noise. The corresponding Euler--Lagrange equations then yield a continuous-time stochastic state-space model (SSM) that describes the system dynamics. The neural network parameters and system states are then jointly learned via a maximum-likelihood method using Gaussian-approximation-based Bayesian filters. The effectiveness of the proposed method is demonstrated on pendulum and Duffing oscillator examples, and its performance is compared with conventional LNNs and with approximate Bayesian filters using known system models.
