STARMAC Quad-Rotor Learns To Do “Provably Safe” Backflips
Writing by Evan Ackerman on Monday, 27 of July , 2009 at 4:57 am
If you had a fancy and presumably expensive quad-rotor UAV, and you wanted to get it to do a backflip, and you were really really smart, you’d probably do what researchers from UC Berkeley and Stanford did with their STARMAC (Stanford Testbed of Autonomous Rotorcraft for Multi-Agent Control) robot: teach it to only do flips if it can provably make a successful recovery.
It’s a fairly simple concept, as illustrated in the graph in the vid: based on the physical characteristics of your robot, you know what it can and can’t recover from in each phase of the maneuver, and you tell it not to perform the maneuver unless it’s within that successful recovery envelope. This sort of logic can be applied to any maneuver (not just extreme stuff like landing on slopes and complex aerobatics), and should help robots make decisions about what they are and are not capable of doing, either autonomously or in concert with a human operator.
Leave a comment
Category: Artificial Intelligence, Research
- Add this post to
- Del.icio.us -
- Digg
No comments yet.