Researchers publish TF-ART taxonomy for force-tactile robot learning
35 authors define a new taxonomy for robots that can feel and regulate force.
A 53-page arXiv survey by Shilin Shan and 32 other authors introduces TF-ART, a unified taxonomy for tactile- and force-aware robot learning. The framework maps how methods organize sensing—including force, tactile, vision, language, and proprioceptive inputs—and builds multi-phase systems that combine high-level policies, action-refinement modules, and low-level controllers. It also covers task settings and infrastructure, offering a practical lens on contact-rich manipulation.
- TF-ART is a 53-page taxonomy authored by 35 researchers across 15 institutions, unifying tactile and force-aware robot learning.
- The framework maps 7+ sensing modalities (force, tactile, vision, language, proprioception) and multi-phase control stacks into a single hierarchy.
- TF-ART provides guidance for designing contact-rich manipulation policies and their real-world infrastructure requirements.
Why It Matters
Standardizes the fragmented field of contact-rich robotics, enabling faster development of robots that can safely manipulate fragile or deformable objects in logistics, healthcare, and manufacturing.