Current News

While my site rebuilding is taking some time, I made a huge leap forward today in the software for auto-soaring. I can now task the Piccolo through Matlab. This means I have the ability to read the Piccolo telemetry (which I have been doing for a while now), do all sorts of fun math (that I have been doing for a longer while), decide what the airplane should do (the control theory part), and then (now!) command the autopilot to do my bidding. It's as if before all I could do was watch the bird, but now I can throw bytes at the bird and it happily will comply.

I have been improving my thermal ID algorithms from the flight data of last Tuesday's 14 flights. I have ID'd several of the thermals my manual pilots were using to get us up to altitude. I also found a few thermals that the aircraft went through while under autopilot control, including the one that I mentioned in the last news post. So ... the missing links are starting to fall into place!

More info and hopefully fixing up the website soon!

Edit 7/06/07: I have fixed all the dead links I can find. Please if you find any more, let me know.

In flight-testing news, I did some gain tuning on Tuesday. Part of the tuning was to fly an orbit around an arbitrary GPS waypoint. By complete accident, I had the airplane orbiting in a location that had a thermal blowing through. Sure as rain, the vario called out lift as the plane orbited around. I haven’t crunched the data yet to get an absolute value, but the aircraft did climb on its own while orbiting. At least now I know I can do auto-soaring by accident, so hopefully by just picking a better point to orbit, I should indeed get the result I’m looking for!

I apologize for the website having several links down ... the site is now updated to version 0.7 that includes an RSS feed! So those of you who would like to be kept updated with any new news postings need only to put: http://goosetech.homelinux.com/soaring/content/rss.xml into your RSS application. I will be updating the broken links over the next few evenings.

Thanks for your patience and I hope you enjoy the new features of version 0.7!

I am very impressed with the Piccolo LT hardware.

I spent Saturday, June 2nd tuning a Piccolo LT to fly a Senior Telemaster with the NC State Aerial Robotics Club. Believe what you will, but after less than an hour air time on the Piccolo, we were shooting autonomous landing approaches! The Piccolo system is much easier to tune than I expected ...

We flew one flight to check the downlink and in-air performance. Then, we flew a flight that tuned the roll-to-aileron gains by prescribing a desired turn rate. For this flight, the airspeed and altitude hold were off, so the aircraft was not commanding elevator or changing throttle. This allowed us to look at the aileron commands and decide that the proportional gain was too high, resulting in oscillations in roll. Once this gain was reduced a couple times, the turns smoothed out. In a similar manner, we worked through the rest of the 7 steps of the gain tuning process.

As I mentioned, after waypoint tracking was working (as in, the aircraft was flying orbits or user-prescribed GPS courses on its own), we decided to just get a feel for how the autonomous landing feature worked, knowing we could wave off at any time. Quite remarkably, the third landing attempt, we just about let the Piccolo LT land by itself! I certainly did not expect after just an hour of tuning that I would be watching the aircraft fly a glide slope and get within about 30 feet of the runway! Had our runway been longer, the Aerial Robotics Club. agreed that we likely would have continued all the way to touchdown.

So, moral of this entry is that I have now completely gone through the process of setting up a Piccolo system to fly an aircraft. I am very impressed with the small size of the Piccolo LT and its performance is equal to, if not better, than its larger counterparts Piccolo II or Piccolo Plus.

The next step for ALOFT is to go through this same tuning process. Aiding me, Michael Allen sent me NASA Dryden's gains for Cloud Swift. It is an identical airframe (albeit with different inertias), so I expect the gains to be very close. I have a scheduled flight June 19th or 20th, so hopefully I will have good news after then. After my good experience with the Piccolo LT, I expect that I shall :-)

One other note before I forget it. We tested lost communications protocol of the system. In response to losing manual override ability, we programmed (for this flight, will be different for competition) the failsafe to default to autopilot mode. If the autopilot communication link failed, we set the aircraft return home and orbit overhead. To test that the failsafe modes worked, we literally turned off the ground station and manual override radios. Like a well-trained puppy, the aircraft realized it was lost and returned home and began orbiting overhead. For ALOFT, this makes me much more confident in the aircraft's response to lost-link situations.

Me in the process of tweaking Piccolo LT gains. You can see the new Dev Interface from Cloud Cap Tech on the lower left of the screen; this is an invaluable real-time tool for monitoring aircraft performance while tuning gains.

News about hardware, software, and future plans this installment...

I recently was given some real copper tape and replaced the aluminum foil that seemed to work wonderfully shielding the elevator and rudder servos. I also learned a neat trick to steal the shielding from scrap coax cable, so the servo lines are now shielded inside the fuselage. I do not know yet if this will increase the RC range with the autopilot on, but I am much more confident I will not see resurgance of the servo jitter I was seeing. Part of me was concerned adding weight with the shielding, but after researching potential single board computers, I think an ounce of RFI prevention is worth a pound of cure.

I decided to move the pitot-static probe to the tail, above the TEK (total energy compensation) nozzle. This way, it is moved away from a potential "strike the ground" location and will also not need to be disconnected every flight test. I repaired the small tear in the wing with epoxy and microballoons (filler).

In other related news, I have been working on better algorithms to find updrafts. My conclusion after a solid week of tinkering is that the flight data I have from the Piccolo does not have good thermal activity. I see several small buoyant bubbles, but nothing organized enough for the new algorithms to latch on to. As a byproduct of working with poor data files, I think my updated thermal locating algorithm is much more robust than the one I developed at Dryden. Fingers crossed that after a set of good flight data, the new methods will ping the center with new precision!

Lastly, I am starting to investigate single board computers for the purpose of hosting the "brains" of the soaring algorithms. My current development has been almost exclusively in Matlab, so the easiest solution would be to run Windows XP and Matlab onboard ... however, this is probably not the smartest thing to do. I'm thinking instead I may switch over to C so I can also use the Piccolo SDK (software development kit) to put new waypoints back into the Piccolo. I think it's time to start the sponsorship search again...

More updates in a few weeks!

Edit: One update already. I'm looking at a modified pitot-static probe from Dywer, model 166. I will cut off the probe right after the bend so I don't have to carry the heavy steel, just the light and accurately-machined tip.

I know I said no more flight tests this semester, but my pilot was available for a nice weather day and so we decided to head to Wilson for a flight test.

The first item on the list was to confirm all the range checks over the past two trips out worked. Sure enough, we walked out 112 paces and still had all the manual control we expected. We even checked all the failsafe modes at distance just to be super sure. In the air we did not notice any range issues, so I am now confidently able to assert the range issues are fixed.

The next item on the list was a maneuver called a POPU (pitch-over-pitch-up). I heard the term by Jim Murray at NASA Dryden and I think he wanted such a maneuver done in order to gather data for computing a sink-rate polar. At any rate, that's why ALOFT went through a POPU today. I should be able to extract a vertical descent rate versus horizontal speed curve. This will tell a whole lot about how fast to fly the glider in various conditions.

Another fun item on the list was to fly via the 900MHz RF link the Piccolo uses. We set up the Piccolo inputs to the servos on the secondary input (input B) on the RxMux safety switch. Since we trust the 1 watt link in the Piccolo more so than the approximately 0.3 watt stock RC transmitter, the failsafe was set to use the 900MHz link in case of RC interference. After a bit more testing, we will likely move to the 900MHz link as primary. Anyhow, it worked so there isn't much story here.

In fact the only real story of this whole trip was that launching in calm air or with a very slight tailwind is not fun. The plane just wallows up the winch line like a lost doggy. Three of the six flights had either a stall or a tip stall while going up the line. We experimented around and determined we just couldn't get much more height with this setup today. You simply have to keep enough air speed so launch height suffers. Oh well, everything is in one piece, so kudos to Adam Propst, my pilot.

With that, I leave you for a bit. Not sure when my next entry will be. I'll try to post periodic updates before I pick back up full steam in the fall with classwork on top of soaring work. Until then!

Mother nature decided to throw a "fake hurricane" off the east coast and had severely limited my ability to flight test over the past days. I had in mind Saturday through Wednesday, hoping one of the five days would have lowish winds and sun. Alas, all that I managed was a full-up range test between drizzles.

On that note, Adam and I took ALOFT to a local field for range checks to see if the conductive foil wrap treatment did anything. The standing conclusion is that the servo jitter was solved with the foil. If you are having servo jitter issues, you can even use something as simple as aluminum foil wrapped around the servo and a layer of masking tape. A nicer way is to get copper tape with conductive adhesive like the larger UAV's have. This certainly has quenched all the jitter tendencies in ALOFT.

The other range checking we did was extensive orientation checks ... we pointed the RC antennas toward each other, away from each other, in parallel with each other, and every other possible combination we could make up. In accordance with dipole antenna radiation theory, both antennas pointed straight up gave the best range (since the donut shaped radiation from the two dipoles overlap best this way).

Also, we did extensive ground-interference checking. With the transmitter on the ground, the measured range was approximately half of just raising the transmitter to waist-level. We successfully doubled the range again by holding the transmitter high above our heads. Doing this same testing to the aircraft itself showed similar results.

The system seemed unchanged by Piccolo transmit power or Ground Station transmit power. We even tried moving the aircraft away from the ground station instead of walking the transmitter out.

The conclusion from two hours worth of ground testing is that the 30 pace range we saw last flight test attempt is within the results of ground interference. By raising the aircraft and transmitter, we were able to get ranges of 120 paces consistently. As to be expected in part, the RC range with the Piccolo OFF was twice the range with the Piccolo on. However, since the system was insensitive to transmit power, I conclude that the interference is less RFI and more EMI from having a computer processor onboard.

Possibilities exist in the future for better hardware to decrease EMI problems entirely. For the updated ALOFT installation, I will drop the Safety Switch setup and move exclusively to the Piccolo stock RF link. Also, I can drop some weight by moving to the Piccolo LT instead of the Piccolo Plus. Weight will be more critical as I add computer components. Lastly, I may write my own energy estimator instead of relying on the Skymelody. Lots of details to ponder. I really want to stop worrying about hardware and move back to worrying about software...

Last for this post, I am moving to Washington, DC for the summer, so I do not know how much I will be able to post about my progress. Rest assured I will have updates at some point.

Thanks for reading!

Well it turns out the interference issue was not resolved by finding the ground loop. During standard range-checking prior to flying, the RC antenna-collapsed range was reduced from ~120 paces to ~32 paces. After an extensive check at the field, the Piccolo transmission was found to be at fault. With the Piccolo not transmitting, the interference is not present and RC range is acceptable.

Further ground testing has yielded a few solutions:

I decided to not fly last Friday like I advertised. Instead, I did hardware and software work at home.

After some searching for the cause of the odd servo jitter, I found a ground loop between the Piccolo, the batteries, and the safety switch. Honestly I should have known because I have fixed this problem before with the Aerial Robotics Club (www.ncsu.edu/stud_orgs/ar), but it just slipped past. With the ground loop fixed, I have rock-solid servos.

I also did some work with the Piccolo Communications document and figured out how to parse Piccolo telemetry data. Currently I can read the telemetry stream into my laptop from the ground station. Moving this code onboard is the next major step. Before I can do that, I plan to do some testing with the saved .tel file from the last set of major soaring flights. This will allow me to test in faked real-time the soaring algorithm thermal locator, which should super fast-forward development and the integration process. However, I won't be able to command the aircraft to a new position and have it respond without a good simulator ... does anyone want to volunteer to add thermals to Flight Gear??

For now, I am planning to fly on Monday April 30th. Check back for more info then!