Shared publicly  - 
 
ARMAR Robot Learns by Watching

A robot helping in the household no longer is a dream of the future. ARMAR, the humanoid robot, can understand commands and execute them independently. For instance, it gets the milk out of the fridge. Thanks to cameras and sensors, it orients itself in the room, recognizes objects, and grasps them with the necessary sensitivity. Additionally, it reacts to gestures and learns by watching a human colleague how to empty a dishwasher or clean the counter.
39
23
Danijel Gajan's profile photoJack Ömer's profile photoDana Venzor's profile photoAndrey Morozov's profile photo
11 comments
 
Cool, thanks for sharing! The learning capabilities are especially interesting in this robot, being able to learn new tricks by observing and mimicking others will make it so much more flexible. I think that in order to get robots to be more useful we need to teach them the social and social learning skills we ourselves use to be able to adapt to new situations.
 
i So need 1 of these's...to carry my laundry..up & down 3 flights!...kewl
 
This brings up the issue, common in a household setting, of learning through observation when the role model has different standards than "best practice" and how the learner gets to best practice.

For instance, assume the role model for picking up stray items in a room was a typical teenager. The robot learner would presumably imprint a standard that wasn't the same as a stricter adult might want.
 
so you mean the robot would throw the clothes under the bed too.lol
 
Well, I'm not being entirely facetious, but I picked an easy to relate to case.

Say a robot learned the standard from a role model that was a malfunctioning robot and unable to maintain the imprinted best practice standard. Do we have auditor robots that check the work? How do we adjust the learned standard?
 
for the robot to work well in a home...first it would have to learn the correct way..so yes i understand what you mean...but then we should be able to show it the way we would like it to be done also
 
it can be trained before it's delivered to a customer's house. it'd be practical if it's possible to turn off the learning ability after training.
 
+Jed Kim Seems as if that would be a very limited subset of typical tasks. For instance, getting milk out of the fridge involves approaching, opening the door, taking the container out, closing the door, and putting the container somewhere.

You could train on images of different containers and even what refrigerator variants look like, but navigating in the typical kitchen to the refrigerator and putting the container down in a suitable spot sounds hard. (I'm happy to be told that these functions are in today's expected scope.)

Finding the opaque, labelled container inside the refrigerator or discarding it when empty doesn't sound too hard
 
for some reason i keep thinking of the I Robot movie,lol
Add a comment...