Reduced-order Neural Modeling with Differentiable Simulation for High-Detail Tactile Perception

TL;DR

This paper introduces a neural simulation framework that combines coarse MPM dynamics with neural decoding to efficiently produce high-detail tactile deformation, improving speed and fidelity over traditional methods.

Contribution

A novel reduced-order neural simulation approach coupling coarse MPM with neural decoding for high-detail, efficient tactile deformation modeling.

Findings

Achieves over 65% faster simulation speed

Uses 40% less memory than existing methods

Improves accuracy in tactile rendering and surface reconstruction by 25%

Abstract

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.