PhySPRING: Structure-Preserving Reduction of Physics-Informed Twins via GNN

2026-05-08

AI Summary

This work addresses the excessive computational cost in spring-mass digital twin models, which often arises from topological redundancy inherited from high-resolution visual reconstructions. To mitigate this, the authors propose an end-to-end differentiable graph neural network that jointly learns multi-level coarsened graph topologies and mechanical parameters, achieving structure-preserving complexity reduction. A key innovation is a dynamic-response-driven node merging strategy that automatically generates hierarchical simplified models while retaining an explicit spring-mass structure. Evaluated on the PhysTwin benchmark, the method improves both reconstruction and prediction accuracy, with the simplified models achieving up to a 2.30× speedup. Furthermore, in Real2Sim robotic policy evaluation, it maintains manipulation success rates and action sampling efficiency.

Abstract

Physics-based digital twins aim to predict the dynamics of real-world objects under interaction, enabling real-to-sim-to-real applications in robotics. Current approaches reconstruct such twins as explicit physical models (such as spring-mass systems) to predict the dynamics, but the resulting models often inherit the resolution of the visual reconstruction rather than being reduced to the physical complexity required to reproduce task-relevant dynamics. This mismatch introduces redundant topology, making repeated forward-dynamics rollouts unnecessarily expensive. To address this challenge, we present PhySPRING, an fully differentiable GNN-based method to reduce complexity in spring--mass digital twins. PhySPRING jointly learns a hierarchy of coarsened graph topologies and their mechanical parameters from observations. At each reduction level, PhySPRING merges nodes with similar learned dynamic responses to optimize the topology, while maintaining every reduced layer as an explicit spring--mass system. On the PhysTwin benchmark, PhySPRING improves dense reconstruction and prediction accuracy over PhysTwin, while reduced models retain stable physical and visual fidelity with up to a 2.30 times speed-up. We further demonstrate the effectiveness of PhySPRING in a Real2Sim robot policy-evaluation pipeline, where the reduced models are substituted zero-shot into ACT and $π_0$ evaluations, maintaining comparable manipulation success rates across downsampling levels while improving action-sampling effectiveness. Together, PhySPRING enables efficient and structure-preserving spring--mass reduction without sacrificing fidelity or robotic utility.

Problem

• digital twins
• spring-mass systems
• model reduction
• physics-based simulation
• topology redundancy

Innovation

• structure-preserving reduction
• physics-informed digital twins
• graph neural networks
• spring-mass systems
• differentiable coarsening