Autonomous discovery of abstract concepts by a robot thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Autonomous discovery of abstract concepts by a robot

Published on May 03, 20113432 Views

Related categories

Chapter list

Autonomous discovery of abstract concepts by a robot00:00
This talk00:01
Scientific goal00:50
Levels of robot programming01:01
This paper02:57
Xpero project03:57
What is "insight"?04:41
"Insight"05:00
Examples of insights05:30
Insights in evolution of theories07:34
Note: This work is not a typical robotics project08:45
Experimental scenario10:15
Experimental loop10:47
Ambition12:04
Insights gained by robots in EXPERO12:26
Example: red object experimental scenario12:59
Some demos, simulated and real robots13:26
Machine Learning methods used20:40
ML methods not used22:09
Learning qualitative descriptions22:59
Object disappearing from camera view24:11
QUIN's Qualitative Tree for Area24:51
Comment25:44
Assessment, CTD.26:11
Learning in Logic, Inductive Logic Programming (ILP)26:39
ILP program used27:42
ILP enables predicate invention28:08
Russell and Norvig 201029:19
Why new predicates: Example30:20
Learning about graspable - 132:31
Learning about graspable - 233:08
Learning about graspable - 334:13
Induced definition35:47
Induced def. in prolog37:25
A scenario where insight happens through predicate invention - 137:55
A scenario where insight happens through predicate invention - 238:00
Robot's prior knowledge - 138:04
Robot's prior knowledge - - 238:07
Negative examples39:08
An induced theory about object pushing39:10
Insight: notion of movable object!40:05
Relative movability40:08
Discovery of concept of "obstacle"40:10
Learning about pushing a "train" ob blocks41:03
Learning about functions of objects41:09
Plans as sequences of pairs41:30
Unsupervised learning41:32
A example of tool use41:33
A plan to achieve the goal42:09
Finding useful patterns in plans42:38
Learning the tool concept using Hyper42:45
Learned descriptions of best patterns43:28
Learning to characterise best patterns43:30
Learned theory with invented predicate P43:31
Invented predicate is a notion of a tool43:36
Conclusions45:33
Acknowledgements45:48