Search papers, labs, and topics across Lattice.
This paper introduces a reduced-order neural simulation framework for tactile perception that combines coarse-grained Material Point Method (MPM) dynamics with an implicit neural decoder to reconstruct high-resolution tactile details. The framework learns a continuous deformation manifold from paired high- and low-resolution simulations, enabling physically consistent and differentiable inference. Experiments show a 65% speedup and 40% memory reduction compared to TacIPC, with improved geometric fidelity and accuracy in tactile rendering and 3D surface reconstruction.
Get high-fidelity tactile simulations with 65% speedup and 40% less memory by combining coarse physics with neural implicit reconstruction.
Tactile perception is key to dexterous manipulation, yet simulating high-resolution elastomer deformation remains computationally prohibitive. Finite element methods (FEM) deliver high fidelity but demand costly remeshing, while Material Point Methods (MPM) suffer from heavy particle-memory tradeoffs. We propose a {reduced-order neural simulation framework} that couples coarse-grained MPM dynamics with an implicit neural decoder to reconstruct sub-particle tactile details from compact latent states. The framework learns a continuous deformation manifold from paired high- and low-resolution simulations, enabling physically consistent, differentiable inference. Compared to the TacIPC, our method achieves over 65\% faster simulation and {40\% lower memory usage}, while maintaining better geometric fidelity. In tactile rendering and 3D surface reconstruction, our methods further improve accuracy by 25\% and produce realistic depth images and surface mesh within a faster inference speed. These results demonstrate that the proposed reduced-order neural model enables high-detail, physically grounded tactile simulation with substantial efficiency gains for robotic interaction and optimization.
