Viam unveils research robot that learns to open boxes from 125 demonstrations
On September 22, Viam announced BoxBot, a research robot for opening cardboard boxes. Visual feedback guides its tape cutter; a model trained on human demonstrations opens the flaps.
Viam says two engineers gathered 125 teleoperated demonstrations in about five hours and run the learned policy on the robot's Jetson Orin Nano. A conference demonstration is planned for late September; warehouse deployment has not been announced.
Our analysis
Divide one job between two strengths
The interesting choice is to split one task between methods. Image-based control handles the narrow tape seam, while a learned policy handles flexible flaps. Joining specialized skills may solve practical problems sooner than expecting one model to master every motion.
If this pattern spreads, people who can decide which steps need precise rules and which benefit from examples will matter more. Robot capability may depend as much on task design as model size.
How could everyday life change?
The following is a possible future based on this news.
A small shop could slowly shrink its pile of boxes
Imagine a small shop receiving deliveries each morning. Staff place boxes; a robot cuts tape and opens flaps while people check the goods and stock shelves. That division could shorten the work left after closing.
This is a scenario inspired by a research prototype. Box variety and safe operation near staff still need testing before shop use.
