Some things come easy to humans but are vexing problems for machines. You can probably stack groceries in a trunk or set a table with little trouble, but a robot stumbles over the numerous constraints we process without hesitation. A team at MIT has created a new generative AI tool that can more easily solve multi-step 3D manipulation problems, which could finally take packing and organizing out of human hands.
Robots can easily understand how to move objects from one place to another, particularly if they’re identical. With more variety come more constraints. For example, when packing groceries, you don’t want heavy things on top of light things. If you’re setting a table, you want to make sure the fork and knife are next to each other, but they also should be near the plate, and so on. Traditional robot AIs handle these problems sequentially, devising a partial solution for one constraint and checking to see if it violates any other constraints. Just a few rules and most systems sputter to a halt or output poorly optimized solutions.
Recent Comments