Mark Mueller from the Flying Machine Arena website and the IDSC presents on Robohub a new failsafe algorithm that enables a quad to remain under control even after the loss of a propeller. The video shows up a quad that intentionally loose a propeller, with and without the new failsafe algorithm:
You can read the full article with more details and photos here:
Quadrocopter failsafe algorithm: Recovery after propeller loss
(don't forget to take a look at the comments below!)
You can also read Robohub's coverage on Amazon's Prime Air and other Quadcopter related articles.
Comments
What they're showing isn't so much a technological feat as a mathematical one. My guess is that they started with Lee brackets, in order to discover the controllable space, and then depending on how tricky it was to go there they either used a simple linearized model or developed a custom non-linear model for a three-bladed UAV. I wouldn't call it simple, but once they publish the paper my guess is we'll see it isn't particularly complex.
Raff D'Andrea's team is so impressive because they always manage to find these simple mathematical truths that are so powerful. For more like this, check out their blind juggler or balancing cube. And for those in Florence for the CDC next week be sure to attend D'Andrea's plenary session, "Acrobatic Flight". I'm hoping he'll bring the Vicon cameras and do some live demonstrations!
Wow....It's good
I mean, with fast enough processing, and fast enough ESC updating, I think this could be handled quite easily with our existing PID control implementation. If we were running stabilization loop at 490hz, with 490hz ESC's updates, and the AHRS didn't get confused by the spinning, and you were in Alt Hold, the algorithm would already give up yaw control in order to allow it to maintain altitude. Not going to happen on APM, but maybe PX4.
So what exactly novel about what they're doing that makes it so much different than anything else? I guess we'll find out once we see the patent filing. :)
Didn't notice that.. That's kinda absurd, like the Wright Brothers patenting aerodynamics..
I want to know what it patentable in the code. Have they patented the idea where the code determines "ZOMG, we're spinning out of control, we must have lost a motor!"
The important part here, is that it demonstrates that it is possible for a quad to fly minus one motor. The robust failure detection etc. comes later.
Well, it is made with a BIG computer processing the IR cams in high speed loop. If is really possible to do it just with on-board processing that guys should show it to us, then we could say "Wow, amazing". It is not so really impressive to me because I saw quads doing the impossible using this process on youtube. That process yes, is very impressive. Thank for share with us.
first time i saw it on ted talk in june 2013:
https://www.youtube.com/watch?v=w2itwFJCgFQ&feature=player_deta...
it shouldn't be technically difficult to detect that one of "arms" does not function, "disarm" motor on it and on opposite arm and start rotating. it remains questionable how quickly you have to rotate and whether existing loops in arducopter allow you to keep thing "controllable" (return to home etc), but it sure must be possible that you just slowly descend while rotating - that's much better scenario than playing a rock falling down..
Hopefully the algo can do more than this. Loosing a propeller can be completely ruled out on a copter. Only newbies suffer from that.
I am surprise how many people are amazed at this, I saw this being demonstrated over 6 months ago possibly by the same guy.
I even tested it with a small toy copter, yaw gain set to full and it worked ok to take off and land.
-
1
-
2
of 2 Next