Abstract
Graph-based learned simulators have emerged as a promising approach forsimulating physical systems on unstructured meshes, offering speed andgeneralization across diverse geometries. However, they often struggle withcapturing global phenomena, such as bending or long-range correlations usuallyoccurring in solid mechanics, and suffer from error accumulation over longrollouts due to their reliance on local message passing and direct next-stepprediction. We address these limitations by introducing the RollingDiffusion-Batched Inference Network (ROBIN), a novel learned simulator thatintegrates two key innovations: (i) Rolling Diffusion-Batched Inference (ROBI),a parallelized inference scheme that amortizes the cost of diffusion-basedrefinement across physical time steps by overlapping denoising steps across atemporal window. (ii) A Hierarchical Graph Neural Network built on algebraicmultigrid coarsening, enabling multiscale message passing across different meshresolutions. This architecture, implemented via Algebraic-hierarchical MessagePassing Networks, captures both fine-scale local dynamics and global structuraleffects critical for phenomena like beam bending or multi-body contact. Wevalidate ROBIN on challenging 2D and 3D solid mechanics benchmarks involvinggeometric, material, and contact nonlinearities. ROBIN achievesstate-of-the-art accuracy on all tasks, substantially outperforming existingnext-step learned simulators while reducing inference time by up to an order ofmagnitude compared to standard diffusion simulators.