Context-aware Learned Mesh-based Simulation via Trajectory-Level Meta-Learning

2025-11-07

AI Summary

Existing graph neural simulators (GNSs) rely on single-step predictions and autoregressive rollout, limiting their ability to model temporal context, infer material properties, and produce stable long-horizon predictions due to severe error accumulation. This work proposes M3GN—the first method to formulate mesh-based deformation simulation as a trajectory-level meta-learning problem. M3GN directly generates complete dynamic trajectories via learned motion primitives, eliminating autoregressive error propagation. Built upon MeshGraphNet, it integrates conditional neural processes (CNPs) with meta-learning to enable rapid adaptation to unseen scenarios—including those involving unknown materials. Experiments demonstrate that M3GN significantly outperforms state-of-the-art GNSs in prediction accuracy, achieves several-fold speedup in inference, and enables robust, high-fidelity long-term forecasting.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Search and Optimization: Sampling/Simulation-based Search
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Simulating object deformations is a critical challenge across many scientific domains, including robotics, manufacturing, and structural mechanics. Learned Graph Network Simulators (GNSs) offer a promising alternative to traditional mesh-based physics simulators. Their speed and inherent differentiability make them particularly well suited for applications that require fast and accurate simulations, such as robotic manipulation or manufacturing optimization. However, existing learned simulators typically rely on single-step observations, which limits their ability to exploit temporal context. Without this information, these models fail to infer, e.g., material properties. Further, they rely on auto-regressive rollouts, which quickly accumulate error for long trajectories. We instead frame mesh-based simulation as a trajectory-level meta-learning problem. Using Conditional Neural Processes, our method enables rapid adaptation to new simulation scenarios from limited initial data while capturing their latent simulation properties. We utilize movement primitives to directly predict fast, stable and accurate simulations from a single model call. The resulting approach, Movement-primitive Meta-MeshGraphNet (M3GN), provides higher simulation accuracy at a fraction of the runtime cost compared to state-of-the-art GNSs across several tasks.

Problem

• Learned simulators lack temporal context for material property inference
• Auto-regressive rollouts accumulate errors in long simulation trajectories
• Existing methods require faster adaptation to new simulation scenarios

Innovation

• Uses trajectory-level meta-learning for simulation
• Employs Conditional Neural Processes for adaptation
• Leverages movement primitives for direct prediction