Paper 
Main CodeThree-minute presentation03:00
Conference on Robot Learning · 2026
Equivariant visuotactile learning
Equivariant Visual-Tactile Diffusion Policy for Contact-Rich Manipulation
1 The Chinese University of Hong Kong
2 Southeast University
01 · Overview
Learning contact-rich skills from fewer demonstrations.
VISTA combines vision, touch, and robot state in a geometry-aware policy that transfers learned behavior across spatial configurations.
Abstract
Imitation learning for contact-rich manipulation requires high-quality expert data that is expensive to obtain. This makes learning a spatially robust policy under limited demonstrations a key issue. To address this, we propose VISTA, an equivariant visuotactile diffusion policy for data-efficient contact-rich imitation learning.
VISTA projects visual and tactile observations into spherical tokens, injects tactile contact cues into visual spherical directions through permutation-equivariant spherical fusion, and rotates the fused harmonic representation using the end-effector orientation. The resulting representation conditions an equivariant diffusion policy to predict spatially consistent actions. Extensive simulation and real-world experiments show substantial gains in data efficiency and spatial robustness over strong visuotactile imitation-learning baselines.
Visuo-tactile manipulationEquivariant learningDiffusion policy
80.9%Mean simulation success
+15.9 ppOver the strongest baseline
73.3%Average real-robot success
50Demonstrations per task
02 · Method
Geometry is built into the policy.
VISTA preserves how observations and actions should transform under workspace rotations, helping the policy generalize beyond the poses present in demonstrations.
Lift to the sphere
Egocentric RGB and tactile images become steerable spherical tokens with a shared geometric structure.
Fuse contact and scene cues
Permutation-equivariant fusion lets tactile contact modulate the visual directions that matter for control.
Predict consistent actions
Orientation alignment and an equivariant diffusion decoder produce actions that follow workspace rotations.
Spherical fusion
Touch gains direction through vision.
The same local contact can call for a different world-frame correction depending on the end-effector pose and surrounding geometry. VISTA aligns these cues while preserving equivariance.
03 · Experiments
Spatial robustness in simulation and on hardware.
We evaluate VISTA on eight simulated tasks and three contact-rich real-robot tasks under limited demonstrations.
Simulation
Eight geometry-sensitive tasks
50 demonstrations per task
Main results
VISTA leads across tasks
Data efficiency
Strong results with half the data
Ablations
Each geometric component contributes.
Removing tactile spherical features, changing the fusion strategy, or weakening the group structure reduces performance on geometry-sensitive tasks.
04 · Real robot
Contact-rich manipulation, up close.
Choose a task to inspect successful rollouts. Paired videos keep the external and wrist-camera views synchronized.
Surface contact
VISTA maintains stable contact while following the target erasing trajectory.
Trial 01External view
Trial 02External view
Precision insertion
External and wrist views show how the policy aligns the pencil with the holder under pose variation.
Trial 01Views play in sync
External view
Wrist camera
Trial 02Views play in sync
External view
Wrist camera
Tight insertion
VISTA uses contact feedback to correct alignment as the USB approaches and enters the port.
Trial 01Views play in sync
External view
Wrist camera
Trial 02Views play in sync
External view
Wrist camera
05 · Citation
Build on VISTA.
The archival paper link will be added after the arXiv release.
BibTeX
@inproceedings{wong2026vista,
title = {Equivariant Visual-Tactile Diffusion Policy for
Contact-Rich Manipulation},
author = {Wong, Lik Hang Kenny and Ma, Yiyao and Wei, Xiu-Shen
and Tan, Zelong and Song, Zhuheng and Xie, Dongsheng
and Chen, Kai and Dou, Qi},
booktitle = {Conference on Robot Learning},
year = {2026}
}