Meta FAIR just launched three new tools that give robots human-like touch abilities. Each tool serves a specific purpose in making robots better at sensing and interacting with objects.
First up is Meta Sparsh, the first encoder that works across different tactile sensors and tasks. It learned from over 460,000 touch-related images without human supervision, making it adaptable to many different situations.
Next is Meta Digit 360, an artificial fingertip that can sense touch with human-level precision. With more than 18 different sensing features, it picks up all the subtle details that make touch so important for tasks.
Finally, Meta Digit Plexus ties everything together. It’s a platform that connects different touch sensors on robot hands through a single cable, making data collection and control much simpler.
These tools could help in many fields. Medical teams could use them to improve prosthetic limbs and robot-assisted surgery. Manufacturing plants could handle delicate parts more carefully. Supply chain operations could pack and move items more efficiently.
Meta isn’t keeping this technology locked away. They’re releasing these tools to everyone, along with research papers and code. They’ve also partnered with companies like GelSight and Wonik Robotics to turn these innovations into real products.
They’ve even created PARTNR, a new benchmark for testing how well robots and humans work together on everyday tasks. This helps researchers measure and improve robot performance in real-world situations.
The impact could reach beyond just robotics. This technology could enhance virtual reality experiences too, adding realistic touch feedback to VR training and simulations.
By making these tools available to everyone, Meta aims to speed up progress in robotics and touch sensing across many industries. The next few years should show us exactly how these capabilities change what robots can do.
