Osegi
AI·1 min read

Ant Group's Robbyant open-sources tactile AI model for robots

Robbyant, part of Ant Group, has open-sourced a tactile AI model that lets robots interpret touch signals from different sensor types through one shared representation.

A robotic hand with touch sensors gripping a small object in a laboratory setting.
Image: techinasia.com

Ant Group's Robbyant has open-sourced a tactile AI model designed to help robots process touch signals. The model handles data from different types of sensors by mapping them into a shared representation, according to the announcement.

Why a shared representation matters

Touch sensors for robots come in many forms, and each type produces its own kind of signal. That has made it hard to reuse software and training data from one sensor design on another. A model that translates these varied inputs into a common representation is meant to address that gap, so the same system can work with hardware from different sources.

By releasing the model as open source, Robbyant makes it available to outside developers and research teams. Open access typically lets other groups test the model on their own sensors and robots, and build on it for their own applications.

What it means for robotics

Vision and language have dominated recent progress in AI for robots, while touch has received less attention. Tactile feedback matters for tasks where a camera alone is not enough, such as judging how firmly an object is held. A shared approach to touch data could make that capability easier to add across different robot designs.

The announcement is notable as another sign that Chinese technology groups are pushing into embodied AI, the field that connects AI models with physical machines. Ant Group, best known for its financial technology business, is placing Robbyant in that area.

Details such as benchmark results, supported sensor models and licensing terms were not part of the information available at the time of writing. Developers interested in the model will need to check the release itself for technical specifications and usage conditions.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:ant grouprobbyantroboticstactile aiopen source

More from this day

See all →

More in: AI

See all →