@musictaps: Peter has been doing this as part of his ongoing research at USC, so this represents several years of work, as well as the work of other researchers. Thankfully, all of his code for doing this is open source so that others can build on top of it.
Test
gelaserguy 1 year ago
Excellent.
I too found the 3D visualisations of motion prediction very impressive.
You are able to recognise movements, generalise them into a format that can be reshuffled on the fly to attain a goal.
This is very advanced.
roidroid 2 years ago
Comment removed
Dirtfire 2 years ago
That could change the world!
Q1: The "generalization" function: is that akin to what the cerebellum does? Or do we look elsewhere in the brain for similar routines?
Q2: Have you seen motion sensing clothing that could track human movements and parlay that into robotic movements?
Illimitus 2 years ago
Comment removed
Dirtfire 2 years ago
thats impressive!
me likes it ;)
keep up the good work!
3xegor 2 years ago
congratulations
mateusaubin04 2 years ago 3
@musictaps: Peter has been doing this as part of his ongoing research at USC, so this represents several years of work, as well as the work of other researchers. Thankfully, all of his code for doing this is open source so that others can build on top of it.
WillowGaragevideo 2 years ago
How long did it take to program the robot to generalize what it learns?
musictaps 2 years ago 2
really cool. It was interesting to see the PR2's point of view with the simulation running over top.
MeebleMeeble 2 years ago 3