DOT-Sim: Differentiable Optical Tactile Simulation with Precise Real-to-Sim Physical Calibration

2026-04-29

AI Summary

This work addresses the challenge of achieving physically accurate simulation of optical tactile sensors, which exhibit high deformability and complex optical responses that are difficult to model. The authors propose DOT-Sim, the first framework to employ differentiable Material Point Method (MPM) for elastodynamic modeling of soft tactile sensors, enabling accurate simulation of large-scale nonlinear deformations. By calibrating the optical response with minimal real-world data and learning residual images to efficiently approximate visual outputs, DOT-Sim achieves real-to-sim alignment within minutes. In zero-shot transfer settings, the method attains 85% accuracy in complex object classification and 90% accuracy in tumor-type detection, while maintaining a mean trajectory tracking error below 0.9 mm.

Abstract

Simulating optical tactile sensors presents significant challenges due to their high deformability and intricate optical properties. To address these issues and enable a physically accurate simulation, we propose DOT-Sim: Differentiable Optical Tactile Simulation. Unlike prior simulators that rely on simplified models of deformable sensors, DOT-Sim accurately captures the physical behavior of soft sensors by modeling them as elastic materials using the Material Point Method (MPM). DOT-Sim enables rapid calibration of optical tactile sensor simulation using a small number of demonstrations within minutes, which is substantially faster than existing methods. Compared to current baselines, our approach supports much larger and non-linear deformations. To handle the optical aspect, we propose a novel approach to simulating optical responses by learning a residual image relative to the real-world idle state. We validate the physical and visual realism of our method through a series of zero-shot sim-to-real tasks. Our experiments show that DOT-Sim (1) accurately replicates the physical dynamics of a DenseTact optical tactile sensor in reality, (2) generates realistic optical outputs in contact-rich scenarios, (3) enables direct deployment of simulation-trained classifiers in the real world, achieving 85% classification accuracy on challenging objects and 90% accuracy in embedded tumor-type detection, and (4) allows precise trajectory following with a policy trained from demonstrations in simulation, with an average error of less than 0.9 mm.

Problem

• optical tactile simulation
• physical calibration
• soft sensor deformation
• real-to-sim transfer
• optical response modeling

Innovation

• Differentiable Simulation
• Material Point Method
• Optical Tactile Sensing
• Real-to-Sim Calibration
• Residual Image Learning