UMI-style robot learning with mobile manipulation and multiple robots

There is a definitive fascination in watching those robots work

This is huge for scaling robot learning outside the lab—being able to run multi-robot mobile manipulation in the real world is exactly what we need to close the sim-to-real gap faster.

the bottleneck isn't learning, it's getting the human demos. does multiple robots help there?

Impressive work. As these systems become more capable, high-quality real-world data seems increasingly important for robust generalization.

Yes, and regular box/sealer/labeling machines do this everyday at 120 boxes per minute. The level of stupidity we are seeing is truly over the top. It’s like a bunch of AI researchers have zero understanding that customized robotics automation have existed for decades now.

sounds interesting, collaborative robotics is the future!