New graph network infers unobserved forces from motion data

Researchers have developed Newmark-η-DGN, a novel graph neural network framework designed to infer unobserved forces and mechanical responses from discretely sampled trajectories, particularly at coarse time scales. This method combines a semi-implicit update inspired by the Newmark-η method with an operator-weighted virtual hub for system-wide coupling. The framework has demonstrated effectiveness in predicting long-horizon motion for systems like deformable beams, human motion, and protein dynamics, even when traditional learned simulators fail. Notably, Newmark-η-DGN can infer mechanical quantities without direct supervision on forces, moments, or constitutive relations, linking coarse-step prediction to the inference of unobserved mechanical properties. AI

IMPACT This research could enable more accurate simulations and analysis of physical systems by inferring hidden mechanical properties from observed motion.

RANK_REASON Academic paper detailing a new method and its evaluation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• deformable beam
• Newmark-η
• Newmark-η-DGN
• protein dynamics

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →