I know we've been posting a bunch about quadrotors recently, but it's hard not to when they keep doing cool new stuff. This demo comes from Pat Bouffard and Anil Aswani and shows (eventually) a quadrotor catching tossed ping-pong balls starting at about 1:40:
All that other malarkey at the beginning of the vid (you didn't skip over it, did you?) talks about the programming that goes into making sure that this quadrotor, with what I think we can all agree is a fairly small container, can reliably make catches. Essentially, the robot pays special attention to what's physically going on with itself, using experience to compensate for thing like increased lift due to ground effect.
This technique is called LBMPC (that's Learning Based Model Predictive Control), and you can see it in action when the quadrotor needs to move sideways to catch the ball, as it figures in the fact that it's going to drift a little bit after it cancels out its lateral movement. Clever.
So, if Berkeley's quadrotor teams up with this robot, this robot, maybe this robot, and of course these robots, you've got yourself a halfway decent chance at giving any Little League team a run for their juice boxes, and I for one would pay money to see it happen.
[ UC Berkeley EECS ]
Evan Ackerman is a senior editor at IEEE Spectrum. Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.