A new system for customizable 3D touch sensors for robots
The OpenTac ecosystem provides scalable tactile skins and design tools for robot control and learning.
Researchers developed OpenTac, a new sensor and full-stack ecosystem for tactile skins. These skins are highly customizable, scalable, and transferable. They are designed for use in robot control and learning applications.
The system uses a scalable resistive tactile skin with a structured deformable overlay. This overlay converts tangential loading into spatially varying pressure. Taxels are the individual units in the sensor array.
The sensor design tool generates fabrication files from 2D outlines and 3D surfaces. It can even use text descriptions like "design a sensor for a size 9 shoe insole". The ecosystem includes a sensor simulator and fast data acquisition.
Adaptive acquisition allows for kilohertz-rate partial readouts. This method is 19.2X faster than full-array raster scanning on a 1024-taxel skin. It samples 16 taxels per update.
The researchers designed and fabricated more than 30 sensors for different robot embodiments. They tested 3D tactile sensing on four real-world robot embodiments. These tests involved contact-rich feedback control and manipulation policy learning.
Key facts
- OpenTac provides a customizable and scalable ecosystem for 3D tactile sensing.
- The design tool can create sensors from text descriptions like a size 9 shoe insole.
- Adaptive acquisition is 19.2X faster than full-array raster scanning for a 1024-taxel skin.
- More than 30 sensors were designed and fabricated for various robot embodiments.