Search papers, labs, and topics across Lattice.
This paper introduces a novel architecture for learning the dynamics of mechanical systems by leveraging discrete forced Euler-Lagrange equations on Lie groups, utilizing only position data. This method preserves the geometric structure and conservation laws inherent in the systems, making it robust in scenarios where velocity data is either unavailable or unreliable. The framework shows strong performance across both synthetic and real-world datasets, particularly in multibody systems with external control inputs.
Learning dynamics from position data alone not only respects geometric invariants but also excels in noisy environments, paving the way for more robust mechanical system modeling.
We propose an architecture for learning the dynamics of mechanical systems based on discrete forced Euler-Lagrange equations on Lie groups using only position data. By formulating the dynamics directly on manifold-valued configuration spaces, the method naturally respects the geometric structure of the systems and preserves geometric invariants and conservation laws. The reliance on position measurements alone makes the framework applicable in settings where velocity data are unavailable or noisy. The approach extends naturally to multibody systems, accommodates external control inputs, and demonstrates strong performance on both synthetic and real-world datasets.
