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

Published in ICRA, 2026

Yang You, Won Kyung Do, Aiden Swann, Rika Antonova, Monroe Kennedy, Leonidas Guibas

arXiv / PDF / project page / BibTeX

Abstract

DOT-Sim is a differentiable optical tactile simulator that models soft sensors as elastic materials via the Material Point Method (MPM), enabling rapid real-to-sim calibration within minutes and supporting large, non-linear deformations. It simulates optical responses by learning a residual image relative to the real-world idle state, and demonstrates strong zero-shot sim-to-real performance on a DenseTact sensor across object classification, embedded tumor detection, and precise trajectory following.

BibTeX

@inproceedings{you2026dotsim,
  title     = {DOT-Sim: Differentiable Optical Tactile Simulation with Precise Real-to-Sim Physical Calibration},
  author    = {Yang You and Won Kyung Do and Aiden Swann and Rika Antonova and Monroe Kennedy and Leonidas Guibas},
  booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
  year      = {2026},
  eprint    = {2604.27367},
  archivePrefix = {arXiv}
}

Recommended citation: You, Y., Do, W.K., Swann, A., Antonova, R., Kennedy, M., & Guibas, L. (2026). DOT-Sim: Differentiable Optical Tactile Simulation with Precise Real-to-Sim Physical Calibration. ICRA 2026.