Current News

ALOFT has a new distance and air-time record of 39.44mi and 3.5hr!

The team did some testing in Northern California flying alongside 12 other teams flying in the Montague Cross-Country Challenge 2008. By "competing" against the manually controlled planes, we could directly compare the performance of the soaring algorithms to the best XC pilots in the world in the same weather conditions. Several teams even carried GPS loggers on-board and were kind enough to share their flight logs, so we have a very quantitative way to compare performance.

The MCCC is split into three days of different events:

1. Speed task: 2hr minimum, open course after completing the prescribed ~20mi course
2. Distance task: no time limit, open distance on a ~72mi prescribed course
3. Speed task: prescribed ~16mi course

The Montague folks were terrific and quite interested. They did not mind that we were using a computer to fly the mission; most in fact were incredibly supportive! Soaring combines a unique blend of advanced technology and gut instincts, so the other RC pilots met the idea of autonomous soaring with open arms. But, I think the best line I heard was when one team drove back to the airfield and my teammate Chris asked how they did. The gentleman stared at his cell phone while replying: "I'm trying to figure out how to tell my wife that we were beaten by a laptop."

I would like to send a warm thank you to Dean Gradwell of xcsoaring.com for hosting the event. His kindness and love for XC soaring is legendary and certainly still underrated. I am honored to be in his good graces and for allowing Team ALOFT to test at his event.

I would also like to thank the competitors at MCCC 2008. Thanks Dieter Maier, John Elias, Dean Thomas, and all the other folks who contributed ideas and asked good questions.

My crew was nicknamed "Team Conservative" for our efforts to stay safe while flying. This started not just from ensuring all connectors are positive locking types, but by incubating a culture of safety practices for all flight tests. For flying, we always keep the plane within visual range so we can manually override at any time. We keep a manual override switch available to the safety pilot so at any time he can have complete control over the vehicle. We keep actively searching for air-traffic and adhere to the "blue sky principle" to keep visual separation between our line-of-sight to the vehicle and to any other aircraft, which ensures the vehicles can't collide. We monitor battery voltages from the ground too, ensuring we don't lose power. With the installed hardware and culture, I consider our team even safer than most manually flown RC systems.

I fulfilled my goal to not make a fool of myself at Montague 2008. Even more, the auto-soaring algorithms showed extremely promising results and I now have a laundry-list of improvements and tweaks to add. Over the next week I will be posting more information about the flights, such as flight logs, altitude plots, and more stories. Check back soon!

The flight test May 30th was a success! The new hardware performed almost flawlessly ... with one exception.

First, Adam and I took a while getting the JR9303 programming tweaked to his liking, I think mostly because we just had to learn some of the radio's quirks while on the runway. No biggie, just a learning process. The range check for 2.4GHz also required less walking; we ranged out at 79 paces (JR suggests 30 paces is "full range"). The directionality of the 2.4GHz system was clearly visable: I will be rotating one remote receiver 90deg to get better reception along one particular direction. But 79 paces was over 2x the distance JR recommends, so we thought that was worthy of flying.

The first flight was off the winch, nothing really major to discuss here. It was about a 4 minute flight with Adam feeling out the new servos and making sure trims made sense. He cut the flight short because there was no reason for him to work the really low lift when we had an aero-tow ship in waiting! Watch the takeoff video.

The second flight was Adam's first aero-tow. We set the plane up on a small PVC dolly, attached the tow cord, and the tow pilot pulled up to give some tension. I was fairly nervous, but Adam had talked the procedure over with the tow pilot and the only thing left to do was just go for it. Like a seasoned pilot, Adam flew behind the 1/3 scale Cub with ease! No problems at all. We released at about 1200ft AGL. Watch the aero-tow video.

Adam continued to check some trims while I warmed up the Piccolo and made sure all systems looked healthy. Then we started flying some large rectangular boxes to make sure the Piccolo liked the new hardware okay. No real problems to report, just a little gain tweaking here and there attempting to clean up some of the turning behavior. Since the plane is now heavier, we also did some speed runs to figure out the new stall speed, which is 19kts in case you were curious. I set the airspeed control to 18kts and we watched the Piccolo do sequential tip-stalls fighting to maintain 18kts. Then I upped the airspeed and the Piccolo recovered. Nice and easy and pretty low-stress too!

Next came the auto-soaring testing. Everything looked healthy, so I turned soaring mode on and we waited for the plane to bump into a thermal. Sure enough, on about the 2nd leg of the rectangle, ALOFT code found a potential updraft and starting exploiting it. The initial turn into the orbit was a bit odd, but otherwise it was gaining altitude just dandy. We soared off and on tweaking items as needed for about an hour and a half.

Just after landing, we walked over to the plane to pick it up and noticed an odd humming/buzzing sound. Upon closer examination, the left aileron servo was stripped and its motor was free-spinning inside the servo case. Doh! When did that happen? Our best guess is that it happened during the flight sometime since the surface was working while sitting on the dolly and it wasn't working after it landed. This also would explain away the odd turning behavior. Kudos to the Piccolo controller for dealing with a completely dead servo!

So what now? Well I have replaced both aileron servos now (ironically they were the only servos I DIDN'T replace). The SBXC and support equipment has been boxed up for another round of testing at a new location. More info in about three weeks :-)

It is time for sharing the past few weeks of hardware updating.

Among the changes:

After spending a night up until 4am and another 3 hours this morning, I finally have my Windex AVL model typed up. The learning curve for the bodies in the the CCT editor versus the AVL body input file method is a bit confusing. Eventually what worked wasto create AVL body file directly, just like an airfoil is defined with x/y coordinates. Anyhow, it's done now, save for some measuring on the actual airframe to confirm the 3-view's scaling and dimensions.

Here's a couple pictures of the model. There are more in the Windex gallery: click here

3/4 view

Side view

Just thought I'd share a link that is sort of a circular reference ... the website DIYDrones.com picked up the latest news about the 29.4mi autonomous XC flight and posted a report. You can read it here: click me

No news is good news, some new hardware pictures are near...

ALOFT flew a course total of 29.4mi completely hands-off in a 68min course time! This yields an average course speed of 25.9mph! Unless someone can prove to me otherwise, I assert this is a new world's record!

As the event unfolded...

Adam and I are wrapping up our school semesters and finally found a day where an available pickup, the weather, and no pushing homework aligned. Friday's weather was 83F and sunny with only 2-5mph winds and 10% clouds.

We arrived at the field armed with Puta, a small Toyota pickup borrowed from a buddy named Alan. (This was the same truck we did the powered run of the XC course several months ago.) Craig was available to drive and beat us to the field. So we took Craig on a drive around the course, Adam teaching Craig how to go about being an XC driver. Around the course, we saw many dark green fields of grass with some speckled plowed fields throughout. After getting back to the field, we set up the equipment and proceeded to put the plane in the air.

The first flights were quite uneventful and downright disappointing. Perhaps with the low humidity and low wind, the thermals were just not organized. Adam tried valiently to catch just one medium thermal to maintain altitude and break through the low-level stuff. No birds were out flying and the completely blue sky was an ominous omen that the air just wasn't moving up today. But Blipmaps predicted today would be a 400ft/min thermal day at 2pm, so were holding our breath.

Finally around 1:30 we had success catching a single good thermal. Almost immediately, we moved over to auto-soaring and Adam watched as the plane climbed out on its own to over 3000ft. At this point, we decided it was time to get in the truck and move out. The first turn is a roughly 2mi flight over a swamp, but 3000ft is about appropriate to safely make it beyond the swamp even if we found no other thermals. I set the aircraft to track waypoint 0, officially starting the flight around the XC course. The plane crossed the start road at 1:47pm.

Off we darted in the little pickup. The plane had worked itself now to over 4000ft, so we were quite confident starting down the first leg. Adam coordinated with Craig to drive ~45mph in order to get in front and stay in front of the plane (we like looking backwards to the plane). After moving through waypoint 0, the vario was calling mostly sink and dutifully speed to fly let the plane move out and stretch its speed legs. What seemed very quick, we made the first turn at 1:51pm and the plane was cooking. It cruised downwind East for leg 2 also without stopping, blowing down the course and hitting turn 2 at 1:55pm, completing the first 4.0 miles in 8 minutes.

At this point, we were a bit nervous, hoping the plane would be able to find another thermal to gain some of the altitude back. From last flight test, we tweaked values on a curve that defined what strength thermal the plane would stop for at various altitudes. This curve was designed to keep the aircraft over 2500ft and cap out a bit over 4000ft. As the plane descended after turn 2, the minimum strength threshold came down and the ALOFT code saw a small upcoming thermal, then latched into it, and then proceeded to start orbiting. The thermal did get better as the plane got higher, so it stayed above the up-moving minimum threshold as the plane climbed higher. At once again over 4000ft, the thermal finally was mighty strong still, but the moving threshold said it was time to move on, so the plane turned out of the thermal and started along the course again.

This is how it went ... stop and watch the plane thermal, see the soaring mode unlatch, hear the plane leave the lift and start hitting sink, watch the plane start picking up speed and moving quickly to the next thermal, hear the lift getting better and check what the current minimum threshold was, stop and watch the plane thermal.

After completing lap 1, the plane was way in the sky at over 3000ft, so we decided it was safe to commence another lap. And then we completed another lap and realized we were still carrying 3000ft at the end and decided to go for yet another! Batteries were holding steady.

On lap 3 about 1/3 of the way through, the plane went through a monster hole of sink and sped up to between 35 and 45kts airspeed. This was also the downwind leg, so we had to chase forward at 65mph in the truck. The plane lost a whole lot of its altitude and was the lowest yet, around 2300ft. I started having visions of landing out. But then the plane made a little lurch, the vario stopped drooping low tones, and speed to fly slowed up a bit. Almost all of a sudden, we heard the vario start chirping. The plane pulled into an orbit and the vario was happy, very happy. In fact, the energy rate readings I was getting were off-scale high, pegging my meter over 19.7ft/s (6m/s) climb-rate! It only held this rate for several seconds, but now I know I need to up my limit! This thermal put us well back up to 4300ft and we knew we had lap 3 in the bag.

It was also about this time we made a call that we knew we had enough altitude to make the field. So, I hit the "soaring off" button, still leaving speed to fly and other stuff active. This allowed the plane to not worry about stopping to orbit, but created a dolphin-soaring type cruise. A few minutes later, we were back to the start/finish line and crossed the threshold at 2:55pm exactly. Whew!

I need to clarify and state for the record I did not touch the soaring computer this ENTIRE flight except for two times: 1. activate soaring mode; 2. deactivate soaring mode. The ENTIRE flight was on autopilot control, all 29.4 miles. On the Piccolo Command Center, the most I did was remove the hold waypoints as we approached them, because they were not needed for this version of flying the course. This did not affect the aircraft's flying and certainly would not have helped anyway. For the record, 29.4miles was autonomous and fully computer-controlled soaring.

...

Post-flight processing indicates ALOFT latched soaring time of 19.8minutes for a total climb distance of 7906ft (this includes the initial climb, not included in the 68min course time). The total air distance traveled was 56.6mi. The approximate total energy gain was 33.4W-hr at an average rate of 190.8W-hr/hr (90W of which goes to level-cruise). Total autopilot time for the whole flight was 94.3min and total manual time was 4.6min.

...

Okay, that was a lot of typing. I would like to share some pictures and describe them. I have uploaded many more images in the XC gallery. I didn't get terribly many while on the course, for the obvious reason we were busy. I did manage to hop out of the truck once and snap a couple.

This is the altitude trace of the flight. I need to demark the official beginning and end of the course run still, but you get to see the raw version and decide for yourself. Max altitude was just shy of 4500ft.

This is a 3D picture of the altitude around the course. I know it is slightly confusing since we made three laps, but you can pretty clearly see how we traded altitude over distance. You can also see the ground track and note that we drifted downwind and you can clearly see the pattern of the course shape.

This is another 3D picture, but the colored regions are where the aircraft was latched into soaring mode. The warmer colors represent where the aircraft was gaining energy fast while the cool colors are a slower climb rate. Here you can clearly see the difference in time on course and time in thermals.

...

Thanks for reading and please feel free to use the comments button to the lower right of this post if you'd like to get in touch with me. Thanks for everyone's support in this endeavor and I especially thank Adam and Craig for an excellent job well done today!

Fear not, I still have more work to do, so this won't be the last post :-)

Craig, a PhD candidate who has attended the last few flight tests, is working on a flap optimization controller. The purpose of the controller is to open the drag bucket of the wing by using flap camber adjustments for certain pressure coefficients measured on the wing. In a nutshell, by adjusting the flap camber based on these pressure differentials, he is effectively creating a wing with a much wider drag bucket (read: speed range).

Craig working in the SBXC fuselage:

The ALOFT team will be flying Craig's controller on the SBXC in the coming month left of the school semester. In preparation, four pressure ports have been drilled and added into the SBXC left wing near flap mid-span. Tubing runs from the wing into the fuselage where pressure transducers measure the air pressure at the specific locations, report these readings to a micro-controller, and modify flap & elevator commands in real-time.

To address safety, the test controller will be used only on the Input B of the RxMux. This means at any time the manual pilot can override the controller and take full control of the aircraft. The RxMux was specifically installed to allow for this mode. Additionally, the 900MHz Piccolo manual pilot link will be used for flying while the controller is being tested, allowing the pilot's control inputs to be directly measured. The telemetry stream is simultaneously measured, providing mountains of data for Craig to evaluate the system.

I have uploaded some more pictures of a recent wind-tunnel test of the left center wing panel. This test was performed in the North Carolina State University low-speed wind-tunnel. The wing was rigidly mounted across the tunnel test section for near 2-d flow. Allowance for flap deflections and for testing various angles of attack were built into the test fixture. This testing allowed Craig to calibrate the pressure readings of the actual wing airfoil rather than solely relying on CFD predictions.

Showing the flap deflection while mounted in the wind tunnel:

I have a couple videos I need to clip together showing the wing mounted and the tunnel running. I also took a piece of string and tried to demonstrate the air flow pattern when the flap is deflected, just for some flow visualization. The video turned out okay and I will post it here soon.

Thanks for reading and I will have the plane back in the air for the XC attempt soon! I suggest checking back for more posts in a week and a half.

Mother Nature threw about 3.5" of rain in the last 48hr and kept Sunday (today) overcast and drizzly at times. Needless to say, this isn't really XC weather. So, the XC attempt has been postponed until two weekends from now, when I'm back in town. Bummer!

I think I finally fixed the weird thumbnail color aliasing! I just had to set the picture thumbnail maker to auto-detect mode. Doh. Anyhow, the pictures should ALL be nicely colored now. I'm trying to keep pictures going up in the Gallery as the other oft-updated part of the website other than Current News. You may have to click "next" at the bottom of the screen to see more galleries.

I also posted some pictures of recent new hardware: New Hardware gallery shows new batteries and a new radio. These are both designed to give more of a safety net for the long distance and long duration upcoming flights. With 2.4GHz spread spectrum system, we're more confident of reducing the number of radio hits, making for a safer manual override system. With the battery, I'm doubling the total energy capacity for the same weight! This will open up flight times considerably. Using a voltage regulator for the servos from the same battery means we'll only have to keep tabs on one battery, greatly reducing the battery monitoring workload. Go-go-gadget-safety!

Just to put this out there again, you can use your favorite RSS reader (MS Outlook 2007, for example) to get emailed when a News update is made. Just paste this link into the URL for the RSS reader: http://goosetech.homelinux.com/soaring/content/rss.xml

Dan

Man oh man was Friday the 21st a good flight test day. We had three flights for 144 minutes of air time, 110.5 minutes of which was on autopilot and 50.6 minutes of autopilot time was spent actively latched into soaring mode. We didn't break altitude or airspeed records, fortunately I suppose!

Flight 1 was an 8 minute hop, getting back in the air and doing one run of aileron effectiveness tuning. We noticed from flying over spring break that the ailerons were still quite busy, even though we have the aileron effectiveness parameter pretty close to right. So today in the calm straight legs and then especially when getting small bumps of turbulence, we could see fast oscillations and watch the ailerons wiggling senselessly. Flight 1 confirmed the aileron effectiveness was not causing this problem.

Flight 2 saw slight tuning of the outer loop aileron gains, pretty much making the plane track better, getting rid of the over-busy ailerons, and now flying around looks more relaxed and happy. Whew!

Flight 2 also saw ALOFT rev5_1 testing ... a LOT of testing. The nice thing was that we split up two options onto separate buttons: speed to fly and speed ring. Ah ha, I gave away my secret XC weapon now too. We were able to tweak the behavior of the speed ring separately from the speed to fly, which helped tremendously. Turns out that with speed to fly on, the plane was trying to fly too slowly and ending up stalling. This explains the problems with soaring from over spring break too. Now the thermalling speed is 1.2*V_min_sink and both speed ring and speed to fly functions are super happy. We flew approximately half the flight with ALL my functions turned on. We did notice that centering thermals was a bit sloppy, even with the advanced features turned on (this I fixed on Flight 3 by switching to my custom method). All-in-all, great flight at over an hour.

Flight 3 was after eating lunch and charging the autopilot and transmitter batteries; the servo battery was deemed charged enough (and we only have two chargers). Immediately after launch autosoaring stuff was turned on and we watched in awe at the best single climb and probably the best centering the plane has ever done. It was a continuous climb from 250m to slightly over 1100m. I do believe this beats Mike Allen's record of a single thermal climb! Yay for another world record! The climb rate was not quite a record, but 414 ft/min this was very impressive for a mid-March flying day (and matched nicely with the Blipmaps 450ft/min max). After peaking out, the plane started its XC run around the field. Dare I admit the plane didn't find another thermal until at 170m and only gained 75m before losing this likely feeder for the next thermal cycle. Adam took back over at 90m, flew over one particular field, bumped in a thermal, then we turned soaring back on and easily climbed out to 800m on autosoar. We know now some tweaks for down-low thermalling and we have a very good estimate of the sink cycle for today. We tested speed to fly all the way down from altitude and are getting happier with its performance. We estimated back-of-the-napkin that for our 1000m altitude drop, we did most definitely over 7 miles, so we're definitely on par for completing the XC course in a more efficient manner than last September's run. ALOFT is getting the hang of this soaring thing...

We decided to stop Flight 3 at the 45 minute mark because we were watching the servo battery getting back to the 4.80v level that caused trouble last flight test. We're pretty sure the speed divergence last time was caused by the aerodynamic loading on the flaps pushing the flap servos back, so we were more careful this time descending. I have both a 2x capacity battery and higher torque servos on the way, so this problem will be moot in the future. The extra battery capacity will be super awesome to have.

I'm hoping for an XC run in one week, last weekend of March. I have several code tweaks that should get done first, but fortunately none are flight critical. With the performance of rev5_1 today, we definitely could have run the course. I'm aiming to make a ~10mph average course speed improvement, up from 4.8mph in September. The XC world record pace is approximately 25mph for 8 hours of flying. Still a long way to go ... but getting closer! The 25mi out and back in 5 hours would be an average of 10mph, so if I can average 15mph in a closed course, I should be able to manage the out and back run. Maybe I should start looking for a 25mi goal and return course :-)

A random but special thanks to Adam for being pilot. The total flight time for the SBXC is 27.5hr, mostly under Adam's belt. I can't count the number of good saves he has pulled in order to put the plane back into the sky for autopilot adjustments. He jokingly says watching the autopilot fly is relaxing, but I really have to tip my hat and say thanks for keeping my airplane alive and happy. I wouldn't be real far along without him around helping keep me or the airplane from doing something stupid.

Check back in a week for XC updates and definitely by then pictures from today's flying.
Dan

PS: sorry for switching from SI to American units so much this post. I normally try to pick one, but I just got lazy this post ;-)

PPS: I pulled my second 15 minute LSF level II flight with Bubble Dancer and filled in a few more spot landings. Either I'm getting better or Bubble Dancer is finding lift on its own now 8-)