System lets robots identify an object's properties through handling

news.mit.edu

11 pointsmikhael3 comments

If they twist the object in each axis, they should be able to get the center of mass and the inertia tensor. You need that to throw something accurately. A ball and a stick with the same mass have very different moments of inertia. Something with liquid inside has a non-constant inertia tensor. You can get a lot of useful info that way. Info that's useful in package handling.

If you lack that info, you have to handle everything very slowly, or clamp with great force, because you can't work with the dynamics of the object. This is one reason robot picking is so sluggish.

We've replaced a $20 scale with a $1M robot arm and ai. Progress.

You might not understand it, but model predictive control requires a model of not just the robot, but also the payload.

This means that you have to hard code the mass and dynamics of the payload or use an algorithm to determine the payload properties automatically.

It should be blatantly obvious then how useless a robot that can only pick up a very specific object and how useful a robot that can pick up any object is.

Reply on news.ycombinator.com