RSS 2026 Workshop on "7th ViTac Workshop: Visual-Tactile Embodied Intelligence - From High-Fidelity Simulation to Contact-Rich Manipulation"

Introduction Tactile robotics has emerged as a key frontier of embodied intelligence, equipping robots with the physical sense required to perceive and interact in unstructured environments. By integrating visual and tactile sensing, robots are empowered to perform tasks demanding dexterity, adaptability, and a deep understanding of physical interaction. Recently, the field has witnessed a paradigm shift driven by high-fidelity simulation. Advanced simulators capable of rendering realistic visual-tactile feedback are now bridging the sim-to-real gap, accelerating data generation and policy learning for robot grasping and contact-rich manipulation.

This workshop brings together a diverse group of leading experts and early-career researchers to discuss these important topics. It is also timely to ignite a discussion between model-based control and learning-based paradigms in the context of visual-tactile fusion and robot manipulation. By synthesizing these perspectives, we aim to cover recent advancements in the area, identify remaining scientific gaps, and foster new collaborations toward human-level visual-tactile intelligence.

The workshop will be divided into two sessions, with invited talks from leading experts in academia and industry, poster presentations from young researchers, and a panel discussion.

Advances in visual-tactile simulation and sim-to-real transfer
Robot dexterity with visual-tactile fusion: control vs. learning

Sessions will be followed by breaks for refreshments, posters, and live demonstrations to encourage interaction among attendees.

Date: Half-day workshop (exact date TBD), July 13-17, 2026
Location: University of Technology Sydney, Sydney, Australia (room to be confirmed)

Past Workshops:
ViTac2025, ViTac2024, ViTac2023, ViTac2021, ViTac2020, ViTac2019

Invited Speakers

Ireti Akinola
Senior Research Scientist, NVIDIA

Haozhi Qi
Research Scientist, Amazon

Wenzhen Yuan
Assistant Professor, University of Illinois Urbana-Champaign

Harold Soh
Associate Professor, National University of Singapore

Maria Bauza Villalonga
Senior Research Scientist, Google DeepMind

Tentative Schedule

Time	Event
2:00 - 2:05 PM (5 min)	Opening Remarks
Session 1: Advances in Visual-Tactile Simulation and Sim-to-Real Transfer
2:05 - 2:30 PM (25 min)	Speaker 1
2:30 - 2:55 PM (25 min)	Speaker 2
2:55 - 3:25 PM (30 min)	Lightning Poster Presentations (2 mins each)
3:25 - 3:55 PM (30 min)	Coffee Break and Poster Session
Session 2: Robot Dexterity with Visual-Tactile Fusion: Control vs. Learning
3:55 - 4:20 PM (25 min)	Speaker 3
4:20 - 4:45 PM (25 min)	Speaker 4
4:45 - 5:10 PM (25 min)	Speaker 5
5:10 - 5:50 PM (40 min)	Panel Discussion: Future of Visual-Tactile Intelligence
5:50 - 6:00 PM (10 min)	Best Paper Award and Closing Remarks

Discussion Topics

1. High-Fidelity Simulation

Tactile Interpretation: What is the bottleneck for simulation across diverse tactile sensors (vision-based and non-vision-based)? Should simulators render high-fidelity "raw" signals (e.g., RGB images, electrical signal) or interpreted signals (e.g., marker displacement, force vectors, depth maps) to maximize sim-to-real transfer efficiency?
Deformable Physics: What are the bottlenecks in modeling elastic materials, soft-body contacts, and complex friction phenomena (stick-slip, hysteresis) in real-time?
Neural Rendering: Can emerging techniques like Gaussian Splatting replace traditional graphical rendering to generate realistic digital twins for visual-tactile sensing?

2. Sim-to-Real Transfer

The Reality Gap: What specific physical discrepancies (e.g., visual-tactile signal or physics) are the primary blockers preventing zero-shot transfer of tactile policies?
Cross-Sensor Transfer: How can we transfer learned skills between tactile sensors with different sensing principles (e.g., optical vs. magnetic) without retraining from scratch?
Benchmarking Standards: What defines a visual-tactile manipulation policy benchmark? What are the essential datasets, tasks, and evaluation metrics for a benchmark in visual-tactile sensing across simulation and real world?

3. Multimodal Perception & Attention

Dynamic Reweighting: How should a robot dynamically attribute attention between vision and touch in challenging scenarios (e.g., heavy occlusion, poor lighting)?
Unified Representations across Vision and Tactile: What is the latent representation for fusing high-dimensional visual data with sparse, local tactile contact information?

4. Control, Learning & Dexterity

Hybrid Architectures: How can we effectively synergize model-based control (for stability and safety) with learning-based paradigms (for adaptability in unstructured environments)?
Human-Like Reflexes: How do we design control frameworks that decouple high-frequency tactile reflexes (spinal level) from low-frequency visual planning (cortical level)?
Safety & Robustness: In end-to-end learning frameworks, how do we impose hard safety constraints (e.g., "do not crush") when tactile feedback is noisy or ambiguous?
Scalability: How can we scale learned policies with visual-tactile sensing across embodiments (sensors and robots)?

Call for Papers/Demos

Topics of Interest

High-fidelity simulation and deformable physics modeling for visual-tactile sensing
Neural rendering and digital twins for visual-tactile environments
Sim-to-Real transfer strategies and bridging the reality gap in visual-tactile policies
Benchmarking standards, datasets, and evaluation metrics for visual-tactile learning
Multimodal perception, attention mechanisms, and unified representations of vision and touch
Hybrid visual-tactile control architectures combining model-based and learning-based paradigms
Safety, robustness, and reflex-based visual-tactile control in unstructured environments
Cross-sensor transfer and scalability of learned skills across visual-tactile embodiments

Posters and live demonstrations will be selected from a call for extended abstracts (2 pages + references), reviewed by the organisers. The best posters will be invited to give talks at the workshop. All submissions will be reviewed using a double-blind review process. Accepted contributions will be presented during the workshop as posters. Submissions must be sent in PDF, following the RSS conference style (two-columns).

Submission Portal: OpenReview (TBD)
Submission Deadline: TBD
Notification of acceptance: TBD
Camera-ready deadline: TBD
Demo Video: Please upload video demos as a single zip under the supplementary material field in OpenReview.
Dual Submission: Papers to be submitted or in preparation for submission to other major venues in the field are allowed. We also welcome published works as long as explicitly stated at the time of submission.
Visibility: Submissions and reviews will not be public. Only accepted papers will be made public.