Tactile sensing is a key enabling technology for real-world, general-purpose embodied robotic intelligence - it is essential for detecting contact, estimating physical properties, and responding appropriately during real-world interaction. Yet, despite decades of research, robotic touch still lags far behind human capability in sensing resolution, surface coverage, response speed, robustness, and ease of integration. On a more practical level, compared with vision, tactile sensing remains less standardized and less widely used, especially for dexterous manipulation, contact-rich control, and data-efficient robot learning.
This workshop will bring together researchers working on tactile sensor design, contact modeling, simulation, manipulation, robot learning, and multimodal perception to examine how touch can advance robotic dexterity. The topic is especially timely for IROS 2026 because the field is moving beyond standalone sensor development toward integrated tactile pipelines that connect hardware, representation learning, control, and real-world deployment. Emerging directions such as scalable tactile skins, tactile simulation, visuo-tactile learning, tactile foundation models, and sim-to-real transfer are opening new opportunities for robust manipulation systems.The workshop will emphasize tactile sensing systems that support dexterous manipulation, where robots must handle objects through complex, changing contacts involving multiple fingers and joints.
We develop a format aimed at bringing together junior and senior researchers from academia and industry, and to highlight recent advances, open challenges, and future directions at the intersection of tactile sensing, learning, and dexterous manipulation, while fostering discussion on how touch can become a core sensing modality for next-generation robotic systems.
We will explore the following four main objects:
To present state-of-the-art advances in tactile sensing hardware, sensor modeling, and large-area tactile skin design.
To highlight how tactile sensing improves dexterous grasping, in-hand manipulation, contact-rich interaction, and robot adaptation in unstructured environments.
To connect tactile sensing with robot learning, including self-supervised learning, multimodal policy learning, tactile simulation, and sim-to-real transfer.
To identify open research challenges and foster a shared research agenda for tactile sensing as a core component of embodied intelligence.
The workshop will combine invited keynote talks, a poster and demo session, short spotlight presentations (contributed talks), and a moderated panel discussion/debate. The format is intended to promote discussion across tactile hardware, learning, and manipulation.
To encourage interaction we will include time for Q&A after talks, a coffee-break poster/demo session, and a guided discussion / debate on open problems like tactile simulation, multimodal learning, and scalable sensor design. The panel discussion will be moderated by one of the organizers. Early-career researchers will be encouraged to participate through poster submissions and short contributed talks.
For the poster session and contributed talks, we will circulate a call for papers in IROS format. The workshop papers are non-archival, and we encourage submissions that have already been submitted to or accepted by other venues. We will host a poster session featuring up to 20 posters, planned to be held in the workshop room. We will select 4 papers from the submissions, for 10-min. contributed talks. We will recruit external reviewers for evaluating each of the submissions.