[Tyler]

Human insights inspire solutions for household robots
5 February 2015, 7:12 pm

People typically consider doing the laundry to be a boring chore. But laundry is far from boring for artificial intelligence researchers. To AI experts, programming a robot to do the laundry represents a challenging planning problem because current sensing and manipulation technology is not good enough to identify precisely the number of clothing pieces that are in a pile and the number that are picked up with each grasp. People can easily cope with this type of uncertainty and come up with a simple plan. But roboticists for decades have struggled to design an autonomous system able to do what we do so casually--clean our clothes.

Source: Artificial Intelligence News -- ScienceDaily

[ivan.moony]

My sister works in garment industry and she says that the most dumb task like picking up just one piece of canvas from a heap is still undoable for machines, so they have to use humans at that place in factories.

[Ultron]

Could be solved if this robot would be enabled to identify how much sleeves are hanging from his arm or how many different textiles, drawn patterns etc. For a household robot I would program it to recognize patterns and later parts of them (striped shirt, checkered etc.) and after it 'learns' all my shirts it could recognizes how many it is holding - judging by how many familiar patterns it notices.

This is not AI however - it is just complex programming. A true AGI would first need to learn so many things before it can do the laundry, and it would also have to think much like a human in able to quickly identify different garments by using a wide range of methods (of computing the data) like we do.

It is also worth noting that not even we can always accurately determine how many cloths we are holding and which. And I guess we can relate this problem to general visual recognition - a similar scenario would be a big shadow dropped by multiple people and the program has to derive an estimate of how many people there is. This would mean using multiple functions - counting heads, extremities and even identifying positions (shapes) of the previous are not possible.

Point is, this is highly complex and they need to teach their robots to walk first, then run.