Rigid Body Dynamics Simulation Based on GNNs with Constraints

Hang Gao, Sheng Yang, WangTie, Fengge Wu, Changwen Zheng

Keywords: Graph Neural Network; Dynamics Simulation

Abstract:

In recent years, the utilization of Graph Neural Network (GNN)-based methods for simulating complex physical systems has opened new avenues for the fields of computational science and engineering. Despite their success, current GNN-based methods for rigid body dynamic simulation are constrained to relatively simple scenarios, hindering their practical use in industrial settings where complex mechanical structures and interconnected components prevail. These methods face challenges in handling intricate force relationships within rigid bodies, primarily due to the difficulty in obtaining force-related data for objects in industrial environments. To address this, we propose a novel constraint-guided method that incorporates force analysis into GNN-based simulations. The model incorporates computations related to both contact and non-contact forces into the prediction process. Additionally, it imposes physical constraints on the prediction process based on Kane's equations. We have rigorously demonstrated the model's rationality and effectiveness with thorough theoretical demonstration and empirical analysis. \textit{Codes and anonymous links to the datasets are available in the supplementary materials.

Primary Area: learning on graphs and other geometries & topologies

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1977