Learning Based MPC on a Quadrotor

Learning Based Model Predictive Control (LBMPC) identifies physical phenomena like the ground effect and uses this to refine the model of a quadrotor helicopter, which improves flight peformance while providing robustness and safety. This enables precise movements necessary for tasks such as catching a ball.

Researchers:
Anil Aswani, Patrick Bouffard, and Claire Tomlin
Electrical Engineering and Computer Sciences
University of California, Berkeley

For more information: http://www.eecs.berkeley.edu/~aaswani/LBMPC

Category:

Science & Technology

Tags:

• quadrotor
• MPC
• LBMPC
• control
• Berkeley
• learning
• robotics

[paronfisk]

I wish I had a trashcan like this, I would never miss.

[NeoTechni]

isnt this how skynet started?

[Nomoreidsleft]

@paronfisk I don't need a trash can like that. I never miss. ;-)

[2osiris1503]

genius.

[cool8man]

This would be awesome to use on a golf range.

Or for Skynet to hunt people.

[eclipsevl]

А наши университеты сварочными аппаратами занимаются и максимум что изучает МорГУ - 16f84 >_<

[clayto1332]

Now this is pretty bad ass

[joshtitle911]

That wasn't a miss. It learned how to throw the ball back at you.

[AVMthrasher]

@BennyTheDirector Today, it catches balls. Imagine what it will do 10 years from now.

[13orrax]

how may times has sarah connor tried to kill you guys?

[3DisFuntastic]

@p6v53as435fe

This sound scary (and annoying) even without a war. :o

[5904454]

That's great!