T3 Contest, Round 2--UPDATE: Deadline extended (and new contest previewed)

[UPDATE: This one proved a bit harder than we thought, so we're going to extend the deadline by two weeks, to 12:00 midnight PST on Sunday, October 18th.

Also, here's a preview of next month's contest: Break the Stanford team's UAV altitude record of 7,142 feet by doing at least 24 circles with a 300ft climb and descent in each! (This won't really beat his official record, because there won't be an official judge there. But you'll get bragging rights, at least). I'll announce full details on Oct 19th]

The first Trust Time Trial (T3) contest was a great success. Lots of entries, nail-biting competition, awesome performances and lots of learning for all.

Now comes round two. The difference this time are as follows:

--Three laps
--3D waypoints. (must hit altitude targets as well as lat/long)

The prize this time is a Global Hawk kit.

Winning entries must be posted in the comments below by midnight PST on Sunday, October 4th 18th.


1) Must complete the pattern as shown above, totally autonomously. Go into autonomous mode before waypoint 1 and stay in for three laps. The four points are arranged in a square, with 200m on a side (obviously the two diagonal paths are longer). Any aircraft/autopilot allowed. It doesn't matter how close to the waypoints you get, as long as you pass on the outside of them.

2) Altitude must be within +-10m of given altitude at each waypoint. It doesn't matter what your altitude is in between waypoints. All altitudes are either above launch position or the contestant's specified "safety altitude".

3) Fastest time to complete three laps and hit the 3D waypoints wins. Must provide GPS track with timestamps and on-board video. (If you don't have/can't afford a small onboard videocamera like the FlyCamOne 2, we'll let it go this time. But in the future: video or it didn't happen!)

GPS tracks are best achieved with an onboard GPS datalogger, like the i-Blue 747 or smaller Sanav ML-7. But if you don't have one or don't want to add the weight, you can just capture the GPS track from your telemetry stream, although you'll have to figure out how to convert it to KML format to export to Google Earth (see below). If your Ground Control System has a built-in map+track function, a screen shot of that is fine, but it should be possible for people to check to confirm that your leg lengths are at least 200m.

Evidence data should include these four things:

1) Total time, along with aircraft and autopilot used. A photo of the aircraft would be nice.

2) Screen capture of path exported to Google Earth or an equivalent, annotated with waypoints and where autonomy began and ended. :

3) GPS datalog file, any format

4) Onboard video, embedded from YouTube or Vimeo. [Not absolutely required but requested]

Views: 2204

Comment by Andrus Kangro on October 9, 2009 at 11:34am
Hi Bill, I am pretty sure that we use DCM in different way. What I do is that I have just one vector for position, pointing up, and I rotate it based on gyro signals (I do not use sine and cosine functions btw, not even in table form, as for small angles, sin(x)/x -> 1 and I use gyro angle itself for sine value - its small enough when measured 50 times/sec. Also, sinx*sinx + cosx*cosx=1, so getting cosine requires taking square root from 1-sinx*sinx).
In addition to that, I have desired direction vector, where I calculate desired pitch angle based on altitude I want, and I calculate bank angle based on airspeed&desired rotation during next 1 sec.
Then, I calculate control vector every second, where I calculate pitch based on altitude change & accelerometer and roll based on pitch&yaw gyro signals + airspeed + accelerometer. I guide plane based on desired&actual vector difference and I feedback control vector to actual DCM vector. So, in fact, I control pitch and roll with DCM, but direction itself only indirectly...
For direction, I use GPS, but I calculate GPS direction from its current and last coordinates - I feel that lag is smaller this way. During this 1 sec which is between GPS updates, I track direction with pitch&yaw gyro - probably should use DCM there, but I feel that more important for prescision is to implement wind tracking into AP. Based on difference between desired & actual direction I calculate desired bank angle. This side I do not understand fully yet - in order to get good results, I needed to reduce allowed turn rate when close to desired direction, otherwise my autopilot did overshoot in pretty much same way like in yours.

Comment by Mark Griffin on October 9, 2009 at 2:13pm
I have an initial entry to submit. Start time: 11:38, finish 17:28. Time=350 seconds.

Setup consisted of an EasyStar with a Paparazzi TWOG v1 autopilot.

I placed a screenshot video of the GCS & camera footage at: I accidentally unplugged the video cable during some of the course so some unfortunate interruption but you get the idea. Video antenna was not optimal either. EasyStar followed the pattern ok but roll and nav loops need to be optimised and with a faster throttle to reduce the course time. Taken near Geneva, Switzerland (you can see Lake Geneva in the background).

The height profile is here:

I had a safety height of 50m so the start is 150m above ground (within 400 feet). Followed the profile quite well considering height information cam only from GPS.
The KML is here: EasyStar_MG.kml

Just my first attempt and certainly won't get 1st place. Unfortunately, I'll be away next week so won't get to do a second more optimal flight. But, if the time limit was extended another week...... I could submit another entry with the EasyStar as well as with my paparazzi equipped FunJet. I also have a Mikrokopter and a PERL based GCS which is capable of following the course autonomously.

Geneva, Switzerland

Comment by Mark Griffin on October 9, 2009 at 2:15pm
The above video URL didn't work. Try this: http://vimeo.com/6986589

Comment by Andrus Kangro on October 9, 2009 at 11:53pm
Mark, when I click on video link then It says that its private video, cannot see it.

Comment by Mark Griffin on October 10, 2009 at 12:21am
I have changed the privacy settings (default was private). Everyone should be able to see it now. Cheers, Mark

Comment by Gary Mortimer on October 11, 2009 at 2:15am

please check for accuracy and flags ;-)

Comment by Brian Wolfe on October 12, 2009 at 2:23am

This is not an entry yet, just wanted to show progress. Between work, weather and all the bug smashing that goes along with bringing a scratch built autopilot to life it's been slow going. I've got the Lat, Lon accuracy pretty good, but in this flight I just kept climbing. A little more tweeking and I'll have an entry in another day or two. This was done with my "Clouds Fly" Easy Star clone. I broke the 300 sec mark but can't touch Vassilis or Andrus yet.

Comment by Gary Mortimer on October 12, 2009 at 2:44am
Well done, thats pretty close

Comment by Andrus Kangro on October 12, 2009 at 9:03am
Brian, you may want to add some helper waypoints, which get you to pass T3 waypoints from correct side. Below is setup that I used for my attempt (its upside down). Regards, Andrus

Comment by Doug Weibel on October 12, 2009 at 10:39am
The deadline is approaching and the weather is poor. However, I might get an attempt or two in. I am way behind the curve and have no onboard camera/video or logger. Too late to order one for T3 Round 2. What is the easiest way to log the flight track using the ArduPilot data downlink? I see there is some logging provision in Labview, but I haven't had any luck using it yet.


You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service