Its been a while so lets try again. Had a hangout with Chris and we briefly spoke about the next T3 round. Chris will work on a top prize ;-)
The mission is simple, get airborne climb to 20m
Fly a cube with 50m sides, pausing for one minute at each corner, so your flight time cannot be anything less than 8 minutes..... Bonus points if you can stay longer at each corner....
The neatest cube KML wins.
If you can do this with a 3D aircraft I feel you would have a very strong chance of winning!
I will close the competition on April the 14th. I will be looking at where you are flying very closely. Please don't try this in public places or within 500m of any building / road.
Comments
Randy, that is looking very good! Do you know what is with the kinks in the two horizontal legs? Did the quad really do that or is it just data?
Bill, is there any smoothing of the GPS data before you log it, either in the GPS itself of your code? I just ask because those lines are so *smooth*. If you look at Randy's latest entry, his lines are quite straight, but they are not smooth. There's a fair bit of low-amplitude, high-frequency "noise", especially on the vertical legs. I don't really think his quad moved like this, I daresay it's impossible. I believe this to be translational noise in the GPS signal as the quad essentially held position but travelled vertically.
This is to be expected I think, as you can see noise in the position estimate if the GPS is parked on the ground. When moving horizontally, the GPS typically does some smoothing. Maybe you've got different settings or something? Just making sure we're comparing apples to apples.
At the end of the day, BOTH groups should be extremely proud of the fact that we're at the point we're sitting here nitpicking details of exactly how the data was collected. When the flight profiles are so good that we question our ability to measure the accuracy of them, that's simply amazing. Compare this to just 1 year ago where we had NO entries in these competitions, and now several of us are duking it out!
Talking to the guys running the SparkFun competition, they questioned whether we could actually fly a copter through a 20x20 meter box (the wicket) based solely on GPS coordinates. I think it's entirely possible, actually quite easy. They figured you'd need a vision system to do it.
As stated, I initially thought the AQ entry was faked, but I don't think that's true anymore. Just reading the AQ thread, you can see that this really happened. If not, it would be an elaborate hoax that doesn't in any way justify the prize.
Gary:
I'm still planning on trying to run the mission with a heli, but it wouldn't be an official entry. I'll probably do a separate blog post about it. It's mostly just to demonstrate the capability of the system on a heli. Final nail in the coffin of the DJI Ace One. ;)
When I saw this contest posted, my next post was on the Autoquad Forum that said DIY has just posted a contest for the AQ.
Their was no doubt in my mind the AQ, anybody's AQ could put in a superior performance. If you poke around AQ video's you'll see guys flying missions with it 2 meters off the ground, I'm not that confident although a properly calibrated one will not climb or descend when going into forward flight. You could fly the paths at 1 ms or 10 ms with pretty much the same results. As a matter of fact if you look at Larry's he flew fairly fast so he would be able to spend the 60 seconds at each corner. I went at 5ms between WP's, 2.5ms vertically except for the initial climb which was 1ms with the minimum time of 196 seconds.
I would be happy to dropbox any or all of the full binary files of any of the 3 missions. They are large, the 42 minute flight is 193,393 megabytes. You could download the QGC from our forum and look at the data that I used to create the KML file and you could export the KML file if you would like to. You can look at all the sensor data that was logged 200 times a second and compare it's performance to whatever FC board you are using.
Randy's last plot looked good and I don't think my MK would do any better.
Bill didn't like the GPS altitude plot and suggested I also show the Baro plot which is what the copter was trying to follow. My response was, I'm not concerned about it being good enough but it being too good.
Cheers
Steve
@Bill, ok, txs for the clarification. I've run out of ammo for pot shots now. :-).
@Richard, yes, I'm sure it's possible..the kml file is just a text file (the kmz is in binary so slightly more difficult). To be honest, when the first AQ entry came along it was so good I had a quick look at it to see if it was a really obvious fake and it doesn't look like it to me. For example, check out the little squiggle at the corner. I think if someone just faked it they would mess up some details.
Randy: Actually no. Steve presented two 3D plots, both of them showing the actual LAT/LON reported by the GPS module. One was showing the GPS reported altitude and the better one reported the barometer estimated altitude. I'm not sure what Larry posted, but I suspect it was the same.
What you have there is a significant improvement. I'm impressed with the progress.
At the risk of sounding like someone with sour grapes, and I'm not trying to take anything away from the awesome AQ entries but I'd like to point out that, as Rob pointed out in an earlier post, the AQ entries are probably using the controller's estimated position as the basis for the kmz/kml files while the APM entries are coming from the actual GPS and baro altitude. That puts the APM entries at a slight disadvantage because they include the inevitable GPS and baro noise in the readings.
Well blow me down, I have just clicked the link to say that it must be the 14th everywhere in the world now by quite a bit. So I will make the first call of the hammer, competition closing going once.......... Please let me know if you are crunching a video, preparing some pictures or whatever. I will close the comp later today unless I hear otherwise.
This was done with a very small quad - a flamewheel 450 clone with a rather heavy 5300mAh 3S battery.
The full kmz file can be found here. Unedited shaky video can be found here (once it's finished uploading)
Much better inertial navigation control.
-
1
-
2
-
3
-
4
-
5
of 18 Next