Every Quadrotor Needs This Failsafe Software | Hacker News Reader

Every Quadrotor Needs This Failsafe Software

spectrum.ieee.org

130 pointsspectruman51 comments

I got my 3DRobotics IRIS a week ago. I've crashed it 3 times already,gone through 4 replacement propellers, and busted my gimbal. None of my incidents were due to a motor failing, I'm just a horrible pilot. Yesterday evening while flying with the kids, we got too high and the wind drifted it down the neighborhood. I switched it into "return to base" mode and it returned itself into an avocado tree.

I guess what I'm saying is: this software looks really cool but I doubt it will solve my current issues :)

Build yourself a foamie and fly that one until you learn RC piloting. Won't help with the avocado tree, but it will surely help with the broken rotors.

http://www.rcgroups.com/forums/showthread.php?t=1666155

Or invest in a simulator package, and learn there.

It looks like Real Flight has quadcopters.

Or buy a small quadcopter such as the Hubsan X4 for a fraction of the price, and practise with that. :)

broken motor/prop is actually very rare (unless you have horrible hardware of course)

still cool (still very much patent pending, too, so also very uncool)

This is pretty cool. TLDR: in the case of failure, the quadcoptor starts to tumble, just like a rocket reentering our atmosphere from a launch. Except unlike with a rocket, the quadcoptor uses the remaining motors to keep itself quasi-stable while tumbling. This keeps the quadcoptor rotating around the path of descent and you can tweak the rotation through the last remaining motors in order to change the over all flight path.

This seems like an effect that would scale with the number of rotors. Makes me wonder about the prospects of a craft with enough small rotors to approximate a sphere.

So its attempting to solve the issue that a quadcoptor can't autorotate to the ground in the event of failure like a helicopter can. Interesting!

I hope the Arducopter community takes note of this algorithm and puts it in future builds. Arducopter is in my opinion the best flight control board for multicopters.

By the way I'm building a community to help RC builders show off their builds. Take a look at some of the vehicles here http://www.rcbinder.com/vehicles . This is my latest build http://www.rcbinder.com/vehicles/aphid-quad

The article says "patent pending", so they'll have to come up with a different one :(

As an aside, the IEEE is in favor of software patents, which is why I'm currently boycotting them and not a member.

To be fair, the IEEE primarily represents researchers (in industry and academia) who spend many years developing unique algorithms that, if any software is deserving of a patent, would be the most deserving. I don't think anyone's in favor of the existence of "swipe to unlock" type software patents.

The owners of the "swipe to unlock" patent certainly are in favor of its existence. Everyone always wants the law to specifically protect them, not those other guys.

Well, just because they've filed for a patent, doesn't mean that it will be granted.

SNORT HAHAHAHA... ok, sorry.

But seriously, even if a patent has been applied for, does that automatically mean everyone else has to stop using it? I don't think that's how it works just based on how easily a process like that could be abused. In reality they should have a few years before it becomes a legit patent right? A lot of things can change in that amount of time.

Not a lawyer, but based on the following resources, one can say...

http://www.uspto.gov/patents/resources/general_info_concerni...

A patent generally does not protect until it has been issued. However:

http://www.irmi.com/expert/articles/2006/warren01.aspx

35 U.S.C. § 154(d) provides "provisional patent rights", namely a "reasonable royalty" after the patent is published (usually 12-18 mos after application) but before it is issued (from 1-30? years). But you do have to have "actual notice" that you are infringing to be infringing, which probably means a notice from the patent applier. (Though if they can prove you watched a YouTube video demonstrating the patent and which said it was patent pending, that'd probably work too.)

So: someone can notify you that you're infringing on a patent in progress, and you have to guess whether or not it will get granted and in what form. Though the answer is unlikely to be "denied", it's quite possible for the claims to mutate tremendously. If you keep going and it gets granted, you've a lawsuit on your hands with liability from date of notice, not date of grant. Fun!

Feel free then, to infringe on any patent that has been described by the applicant but not yet been published by the USPTO. After that, you've got until they catch you (though that's a dangerous game: you're only a friendly jury away from trouble, as the law on "notice" seems not entirely settled.)

What a great system, eh?

I've just scanned your post, it appears largely correct. However you're generally allowed to use the teaching of a patent for personal - non-commercial - use, for research and such.

Though in US law it seems non-commercial/experimental isn't a complete mitigation as it is elsewhere.

Patents have not prevented open source implementations of heavily patented audio and video formats. You just need developers and users who are willing to disregard the patents.

Can you really patent the implementation of PID algorithm? There isn't really anything novel about the fact that it is on quadracopter platform as opposed to any other control system other than it seems to be well executed, fast, and effective.

As long as you don't use it commercially you're able to [personally] exploit the teachings of a patent document. It's part of the patent deal.

Edit: a brief check on the current US situation shows it's been eroded dramatically and that, like the copyright contract, the patent contract has become a far weaker deal for the public.

Something like this is already implemented in some systems, like WooKong-M from DJI[1]. I'm not sure if it can handle also quadrotors or if it's a hexarotor implementation only.

[1] https://www.youtube.com/watch?v=cNdzBRNNkAw

Neat site, though some of the links appear to not work. Ex: http://www.rcbinder.com/vehicles/piper-pawnee gives me a generic error message. ("We're sorry, but something went wrong.")

It was pretty lame when my quad caught a gust of wind, flipped upside down, and cut off all motors to plummet ~150' onto a brick wall. Just had to replace the crossbeam, but an ability to recover from extremely harsh conditions would make me far more comfortable with flying it (higher than 15') anywhere but a grass park.

How the fuck would it handle triple loss? 2 engine failures I can understand 3 not so much.

Also hope their patent fails. Physics should not be patentable.

Imagine the thing tumbling. some % of the time, the remaining rotor will be pointed in a useful direction, so you can give it the falling quad a push. If you want it to fall slower, run the rotor when it's pointed vaguely up. If you want it to fall somewhere safe (say to the north), run the rotor when it's pointed vaguely north.

I imagine that, that would mainly leed to the thing starting spinning more and in the end make the crash even worse.

if they could fly a quad copter with only one rotor, the engineering behind that is according yo me good enough to allow a patent.

(I can't watch the video) It certainly seems plausible that it could work.

Imagine a copter with reversible motors that can respond quickly-enough (weasel words!) to inputs from the control system. Now, the copter has angular momentum along the direction perpendicular to the rotor disk and a vector between the copter's center of mass and the working motor, and further that the copter is oriented such that the vector is also parallel to the plane of the ground.

As jfoutz said, running the motor ("forward") when the copter is upright and off otherwise gives you net upward thrust (averaged over time). It also applies a torque to the copter, which will mostly be parallel to the extant angular momentum vector. So, a first-order fix is to do what jfoutz said, but also run the motor "backward" (negative thrust) when the copter is upside-down. However, rotor also generates some torque perpendicular to the rotor plane, which will cause the angular momentum vector to rise above the horizon (since the motor reverses when it is upside-down, the angular momentum vector rotates upward rather than revolving around the original direction of the angular momentum vector).

However, if you allow the copter to have some angular momentum (i.e., you don't try to completely eliminate it), the rise might be slow enough to be managed by running the motor at other times. For example, by running the motor when the rotor plane is perpendicular to the ground, you could create a similar drift in the angular momentum vector. But now it slowly drifts around the compass instead of rising above the horizon. If you make the "compass" drift faster than than the "altitude" drift, the torque from the first one will average out to something very small. So by managing the rate of these two drifts with an appropriate feedback loop, you can probably obtain a stable solution where there is a net upward thrust and the angular momentum vector "slowly" (for possibly large values of slow) precesses along the horizon.

With reversible motor I see no problems it would take a shitload of a motor but it could definitely work.

Unfortionaly the engines of the hobby quad copters we have today are not reversible without a relay. And even if they where they would not be able to change direction fast enough. One thing that could work is pitchble blades, the problem with that is that it introduces a much bigger risk for failures.

I could be wrong, but I believe that the point is that this isn't something that a human pilot could do, but is possible with good software and sensors.

> Physics should not be patentable.

Physics isn't, particular applications of it are. Otherwise, nothing would be patentable. Well, except software.

Math isn't patentable so software should not be patentable.

Their patent isn't physics; it's an algorithm ... which is equally unpatentable.

While we're on the topic of "every quadrotor should ____", how about the improved efficiency of this design?

http://spectrum.ieee.org/automaton/robotics/aerial-robots/ir...