Current News

We're back flying the SBXC again!

It's been almost 3.5months since Montague, and it's nice to be flying the glider. Adam and I drove out to Wilson for what we expected would be a mediocre day. Turns out we had a 45minute flight off the first launch. Then we did a few sink polar flights that were puny little 3minute hops. Then on a sink polar hop, the plane bumped into some lift and we switched modes quickly and turned on soaring. The plane worked some smooth but weak lift a couple times for a flight of just under an hour. Pretty good for a crummy day!

The main focus of the flights was to check out some code changes I made over the past several months. Probably the most fun change was testing the left/right turn direction decision logic. A lot of full-scale pilots ask me how the plane decides to turn left or right initially. Well now I have a better answer than "it always turns left." Instead, since I have the GPS coordinates of the expected thermal and since I know where that coordinate is drifting due to wind, when I latch into a thermal I know if the orbit point is behind me to my left or to my right. When it's to the right for instance, the Piccolo will start that orbit by making a right turn. So, I set a right orbit. Viola, finito! It seemed to work reasonably well yesterday and it was nice seeing the plane have a new behavior. We'll have to evaluate this better by flying an XC run, but the idea looks very promising.

Surprisingly, we had plenty of altitude to work on some flap polar data points. One item we noted at Montague was our total lack of speed range due to not using any flap camber adjustments. So, the next idea is to create a sink polar for several flap settings to open up the speed range, especially for inter-thermal cruise mode. We tested -1.5deg and +3deg flap settings, working on building a flap polar for each of these. Unfortunately, it looks like there was too much noise from bumping through lift or sink to get good data. We really should try the sink polar testing on an early calm morning.

I think we'll have to dedicate a full flight testing day to dialing in the speed control loop a bit more too. The lower airspeeds seemed to capture just fine, within around +-1kt hold, but when commanding for example 40kt, we actually saw an excursion to almost 60kt as the aircraft overshot the commanded speed. This definitely causes some efficiency losses because the plane will lag behind the proper airspeed to fly.

With that, there is more work to do! But it's good to be back with the familiar SBXC again :-)

Thanks for reading.
Dan

Flew the Windex again this morning! After having issues with the winglets on last landing, we decided to leave those off. Also, we flew the 2x battery configuration again, for a total of 4s 16Ah available.

Takeoff was pretty uneventful this time. Keeping the wings better level before letting go helped the plane track a bit straighter. The nose still burrowed, but the grass was damp, so it slid along pretty well before the elevator had enough control power to hold the nose up.

The cloud ceiling was low, so we were only able to fly up to around 750ft before the plane got fuzzy. No big deal, just meant cruising around on power.

Most of the 33 minutes we spent figuring out trim and throttle changes with flaps and/or spoiler at various settings. Batteries ended at 15.03v, so we're about half way down the lipo discharge curve, according to this website. We could probably have squeezed another 15 minutes out of the pack. The approximate usage was 50% throttle except for 30sec of 100% throttle for takeoff and one bump of full throttle before setting up for landing. I think having my third pack in parallel will be nice, but I'm not sure if I can hit the 2hr mark that MotoCalc is touting.

It definitely pitches nose-over with full throttle, but half throttle we have trimmed out okay. With low throttle, it does pitch down a little bit and pick up some speed, but this is also from the fixed prop giving drag and increasing our sink rate. We will be setting up a throttle to elevator mix before next flight.

For flaps, we have a flap-to-elevator mix per Baudis' recommendations, however it still needs a touch of up elevator for the full flap configuration. We could definitely tell that with full flap, it takes a bump extra throttle from mid flap to maintain altitude. From clean to mid, throttle is approximately the same, with perhaps one click less. Mid flap looks great for max endurance cruising.

For spoiler, we found last flight test that the Baudis recommended mix is backwards, perhaps just a linguistic mix-up. With spoiler out, up elevator is needed, NOT down elevator. However, even with flipping the direction, more up elevator is needed for full spoiler than the 3deg recommended.

We landed on a flatter area this time and tried hard to keep the speed up more than with the SBXC. Coming in reasonably fast allowed setting it on the main wheel with no issue and keeping the wings level was quite easy. The plane happily bounced along until the speed bled off. This will be a more typical landing approach for us, modulating spoilers to keep the glide slope and holding speed up higher than a wallowy-almost-stall.

The recommended deflections definitely make the plane pretty spritely and nimble. The rudder doesn't do much for turns, but the aileron to rudder mix works so you can fly right stick only anyhow. Even with more pitch inertia than other Windex configurations, the plane is still pretty pitch sensitive. I'm not sure if this is due to having the CG at the 75mm mark and that being slightly aft or what. It does seem to trigger air motion with the 75mm mark pretty nicely. Obviously there was no soaring with the cloud cover, but it was easy to tell there was still some air motion going around.

Oh yeah, it flies just fine without the winglets :-)

I was curious actually how we placed at Montague, so I figured out the man-on-man scoring for the three days. Man-on-man works by normalizing scores to 1000 for the day's best performer, then giving each competitor a fraction of 1000 points based on their performance fraction. For instance, say John Doe flew for 2hr as the best pilot of the day; he would receive 1000points. Say Jimmy Dude flew 1hr; he would receive (1hr/2hr)*1000 = 500 points.

Using this MOM method and the Montague scores found at xcsoaring.com, the results come out to be:

As you can see, ALOFT Team 364 came in 3rd place overall! Holy moley, we placed!

Dan

We maidened the bird on Sunday with a nice headwind. The Piccolo and antennas weren't installed, so this was around a 16.5lb takeoff weight. I also only had the parts to put two batteries in parallel, so the third rode along for ballast purposes.

Takeoff was mostly uneventful. The roll in grass was a bit difficult looking, trying to keep the tips out of the grass and keeping the nose from pitching over from the high thrust-line. The nose did burrow, but was slick enough to slide and let the speed build. Keeping the tips out of the grass seemed not too terrible once a bit of speed had built up. We only used about half of the runway.

Once in the air, it was a rocket, able to speed up while climbing fairly steep. Typical maiden flight stuff like working trims and getting mixes nailed down was pretty uneventful. We were able to make several climbs back to altitude at around 75% throttle. Pulling back to 0% throttle, the plane really does move across the sky and seemed to kill altitude pretty quickly. It is unclear how much of this is due to the high wing loading or the prop acting as an air brake. I suspect the higher weight after doing a stall test and seeing a severe tip stall. With the flaps down and spoilers up, the stall was much nicer, possibly due to the higher CL of the root section stalling first.

In the 22min flight, we shot several landing approaches to get the speed and the visual lined up. It will slow down, but still comes in fairly fast compared to the SBXC. Adam was able to set the plane down easily enough, but one winglet caught the grass and rotated the plane around, letting the opposite winglet now help arrest the sideways motion. That outer winglet cracked along its seam, but should be easily fixed with some cabosil spooey injected inside. On the SBXC maiden, the rudder hinge broke clean off, so having this cosmetic fix seems typical and no big deal.

I hope to fly again this upcoming weekend. We'll get a better feel forthe soaring performance at this higher weight and see hopefully how tough it's going to be for the soaring code :-)

I uploaded a few select pictures from the flight in the Windex Maiden gallery. Enjoy!
Dan

I apologize for a delay in posting. I have been busy traveling to the AIAA Guidance, Navigation, and Controls conference in Honolulu, Hawaii and then promptly moving back to school thereafter.

Speaking of conference, I have now published a paper! For all you folks who are interested in reading details, I now have a good way to share them: See the first page here. If you would like the full paper and don't want to pay AIAA $25, I will happily email it to you if you send me a short email asking for it. I'm sort of doing this so I can just get a feel for who is interested in this project, so a short little "this is who I am and why I'm interested" would be most appreciated.

Other school news is that I signed up to stay for my PhD instead of stopping this fall at a Master's degree. This means I'll be working on the project for a couple more years and hopefully be able to keep this site updated as I have been doing. We'll see how much interest in the paper there is and that will probably tell me how much effort I should keep putting into this site.

Thanks again for all your support and please let me know if there is anything more you would like to see on this site or any way I could improve it. I am also most interested in feedback when you read the paper for ideas or suggestions!

Email me.

Dan

Just a quick note, turns out that Cloud Cap Technologies listed my autonomous soaring project under their News Section. Thanks CCT for your products!

In sort of related news, I have been working away on updates to auto-soar, such as trying to figure out the "all important" left versus right first turn decision. Also, this will likely be Adam's last semester of piloting because alas he is graduating in December! As a result, we will be working on a record-run this fall before the fall weather starts to stamp out the good soaring weather. Speaking of graduating, I have decided to stick around school to get my PhD in Aerospace Engineering. The auto-soaring project has been morphed from a thesis to a dissertation topic.

Thanks for reading! Stay tuned possibly for a Windex maiden flight in a few weeks!!

The server that hosts this humble little site is getting an upgrade and as it has been going through some growing seg-fault pains. Not only that, but it is going for a car ride and moving! So just to let any regular readers know, if you happen to see the site down, please try back in a day or two. My most gracious sever host gives his apologies and hopes that you have enjoyed his service thus far. Thanks Mr Server man!

In other news, I will be presenting my Auto-Soaring work at the AIAA Guidance, Navigation, and Controls (GNC) conference in mid August. If you happen to be in the Honolulu area, maybe you can drop in! Once the paper officially makes the proceedings, I will post it for your personal review here. Check back about August 25th for that news update :-)

Dan

I realized that about the only big item I haven't done is write up a nice flight report about Montague ... oops!

We arrived on Thursday about 2pm and drove into the Montague Valley, practically in the shadow of Mt Shasta. Being from the East coast and having not ventured into the Northwest before, this was about the prettiest place on earth I could ever imagine flying. We spent Thursday late afternoon putting the plane together, going through checklists and safety checkouts, then put the big SBXC into the air two or three times. This was just to check out that the systems were all healthy and that shipping across the country didn't produce any hidden faults. Flying Thursday was limited to about 400ft, coming off the winch and finding zero lift. But, the airplane and systems checked out happy enough, so that was a success.

After flying Thursday, we talked with folks enough to get an idea of the course and Dean Gradwell gave us the packet of turn point information. About an hour before dusk, we drove most of the course, except the southern Grenada to Weed legs. Having this preparedness really helped settle my nerves about having an out landing. This area has so many fields and open spaces that my worst fears were waning. Still, a lot of fields were littered with rocks and some with very sharp pokey looking bushes, so landing out was very much not desired.

Thursday late evening I spent planning our course. I took Dean Gradwell's GPS turn points and programmed them into the computer. From here I could make some road-following choices and also avoid terrain. From driving the course, I knew we generally wouldn't have any terrain concerns, except for the rolling hills that would make ground level change from airfield altitude. This just meant I had to be careful when quoting altitudes to account for the rising or falling ground elevations.

Friday morning finally showed up and so did the masses! I think around 12 teams were competing, each generally with three or more folks on a team. I saw several patient wives and children hanging around the airfield too. All in all, there must have been 45 people at the first pilot's briefing between teams, family, and a handful of full-scale pilots. The briefing was what you might expect: road rules, contest rules about turns and distances, air-rules, manners, etc. It was easy to tell the CD (whose name slips my mind) has run this contest before. Everything is quite organized and it sure makes things fun since everyone is able to concentrate on flying and having fun rather than remembering complicated rules.

Anyhow, at the pilot's briefing I was awarded a 25km distance pin for the Spring semester's flying in Wilson asked to give an introduction of the auto-soaring group. Most everyone seemed happy to have us onboard and I got several very good and intelligent questions about the algorithm. Nobody seemed skeptical; curious is probably a better word. Very quickly, the questions cut through my past two years of work and I was answering search methodology and flap camber questions, neither of which I have really done work on. Just goes to show that XC pilots are an intelligent bunch of guys!

The pilot's meeting eventually concluded and the winch opened for official contest flights. The day seemed very slow to get started, so some folks flew around the field, but nobody really had a good run until after about 1pm. Friday's task was a ~26mi prescribed course with a 2hr minimum. After completing the course, if you still wanted to fly, the course was pretty much yours for the picking.

Team ALOFT waited a while for several folks to get on course before we pulled our stuff out. Being "team conservative," we didn't want to risk a mid-air or worry about others already on course. So, about 1:30 we finally got in the air and turned the flying over to the Piccolo. Adam kept vigilant watch, minding the other SBXC and MXC in the air to make sure our paths didn't cross. After about 15 minutes of working altitude at the airport, we geared up and hopped into the car and off we went!

The tree ALOFT team members were Chris (driver), Adam (safety pilot), and Dan (back-seat driver). The rented Seabring convertible was turned into a rolling operations center. Some folks had pickup trucks, but we figured we'd be better off all having seatbelts and the ability to stay out of the wind. We also figured a convertible would be quieter for better inter-car communication. Chris and Adam sat up front and I squeezed into the lonely back seat. My mobile ground station pelican case barely fit into the back seat sideways, but that also meant it was held in well. Originally I was trying to use two laptops, but there just wasn't enough room.

Back to the course ... we went zooming onto the road (obeying all traffic laws and stop signs, yes) at a measly 35mph. Sometimes Chris would pull off to the side of the road and have his flashers on so the local traffic could pass us by. There wasn't a whole lot to worry about, but the occasional big-rig came by, so it was really important to be visible. We would stop and I would call out what the aircraft was doing (typically airspeed and altitude calls were what Adam wanted). When we'd catch a thermal, I'd keep the team up to date with thermal strengths, altitude gains, and which way the wind was drifting. The miles slowly ticked by...

Friday was quite an adventure. We had driven the course on Thursday, but it was really different with an airplane in the sky. I didn't feel unsafe or nervous, I just felt eager; eager to catch that next good thermal, eager to get some good altitude, eager to find out what the next challenge was going to be. It's amazing how fast an hour goes by when the flying is exciting!

The conditions on Friday were not what Adam and I had expected. We were ready to see substantial +80fpm climb rates and massive hat-suckers that could take an unsuspecting plane into the ether. Instead, most thermals we caught were mild 1-2m/s climb rates and topped out at 1500-2000ft. For Wilson XC flying, we would have not even bothered hopping in the truck. Here, the thermals would hit some invisible atmospheric barrier and just disappear! Over the weekend a few folks had theories, but nothing helped the thermals get any higher. Occasionally we could work a strong core up to 2500 or 3500ft, but those were rare and we generally only found one a day. Some other teams said they never got above 2000ft the entire weekend. For Team ALOFT, this was a difficult fact. Most of the tuning and tweaking of the ALOFT system has been aimed at keeping the aircraft between 3000 and 4000ft. With the lift topping out considerably lower, this meant the strong arm of the computer code would be practically negated. ALOFT would have to work hard at much lower than usual altitudes.

Another item that we didn't expect about the weather was the spacing between thermals. Adam has some personal obsession with knowing the natural spacing between thermals. The local gurus talk about a "thermal cycle," sort of akin to a natural 5-15minute cycle where the lift is either abundant or nonexistent. Timing can be critical if there isn't much altitude left. At Montague, the spacing between thermals was absolutely enormous. On the East coast, we could get away with flying a straight line between turn points, finding plenty of lift along the way before bumping into a really good thermal. At Montague, the good thermals we found were often 1 to 3 MILES apart. For a glider with a 14 foot wingspan, a 3 mile glide is a looooong way.

So Friday turned out to be a good day in the end. We completed the whole ~26mi course and even saved enough altitude to make the 3 mile trek back to the airport. Yay for not landing off-field!

Saturday morning rolled around and the morning pilot's briefing gave the results from Friday. We placed 6th for the day, but considering the amount of items we learned just from being on the road in this location, I was happy to be in the air making progress. Saturday's task was where we hoped to make a bigger impact. It is a 72mi course. The first 6 are heading North, then some 35 miles south, then traversing back along most of the points back to the airport. In the past, folks were able to complete this mission, but with slow conditions on Friday, folks weren't expecting much. Still, the air was better a bit earlier and we hit the road around noon. The tweaking I did on Friday night was paying off. The airplane was doing a bit better picking which thermals to stop for and the plane wasn't missing quite as many of the small ones. Still, it was an effort to get out of the airport and move to the first waypoint. But, being a distance task and not a speed task, there was no incentive to push. Team Conservative could play slow and try to stay high!

On the first North leg, we did better staying high. The locals said to stay over the valley or over the hills, whichever was producing better. We found good lift over the valley, so the flight plan was shifted off the road. The 6mi south leg back to the main loop was more or less a breeze: downwind and good lift along the way. The plane was up to 3500ft AGL over top of the hump, so we were able to drive alongside for several miles without needing to stop for lift. Then it started to get harder. Having learned where some of the house thermal generators were, the updated flight path did a better job bumping into good air, and since we knew what to expect, it was a lot less stress driving. We did have a few snafus with the latching function and drove through some thermals we probably should have caught, but we pushed on anyhow, stopping for a good boomer at Philippe and again at Oberlin. At this point, we picked up a team from Portland in a matching blue Seabring, and we played leap-frog with them. They would stop for a thermal and ALOFT would cruise up and latch the same one. Maybe we would climb out better and leave first, only to be followed by the Portland team, or vice versa. That was a lot of fun! They even drove by and joked "hey, this is man versus machine!"

Miles and miles went by and we watched our altitude buffer get small. Down near Grenada, we were having trouble getting above 1500ft and were starting to see altitudes below the 800ft mark. These low thermals we have found are often not well organized and are small (read: hard to find when you're going in a straight line). We did our best when it was getting really low to say "the fields on this side of the road look better, so let's move the course to the left." When it's low, it gets hard even for manual pilots to find good air, so we called this pretty even ;-) The car was getting slower as we passed through Gazelle and eventually passed the Portland team as they worked a small bubble about two telephone poles high. These guys were out of their car and had a landing area picked out. The writing was on the wall... We made about 3 more miles and were down to 300ft AGL. Adam was itching and my code didn't seem to find anything, so I reluctantly said "Adam, all yours, see what you can pull out of your hat." He found one small bump, but it wasn't big enough to use, and so he guided to plane to a relieving skid and stop back on terra firma. 207 minutes course time. We zeroed the odometer once Chris pulled back up to the plane, disassembled and put the beast back inside the car, and drove 3.3 miles to the Gazelle checkpoint. It was also at this point that the Portland team came driving by, looking way up to their plane well into the 1500ft range over the hills; they made a great save and found the hills were kicking off enough lift to get some altitude. We headed home having bagged a super awesome 39.44mi. I was floored as we drove back on the highway at 70mph and it still took a while to get back to the airport. Geez 40mi is a LONG way!

Sunday was back to a speed task. The course was short, only ~16mi. Last year, Joe Wurts and company flew a plane to "seriously high" in the 4000 - 5000ft range and ran the whole speed course without stopping. If ALOFT could get up to this altitude, its speed to fly implementation and straight-line flight path ability would certainly have a strong advantage. Alas, the lift still was topping out low. We got up to 2500ft near the airport on the first attempt, but struggled to find any more good lift. I think I was surprised we actually had to land off-field again, this time only ~9mi in to the course. The second attempt was more fruitful and yielded a completion of the course! The highlight certainly was a dramatic "we need lift" followed by a bald eagle showing us the best thermal of the weekend. We caught up to another team that followed the eagle into the same thermal, except they left early and we waited until topping over 3200ft. Since I have no final glide calculator, we made an educated guess when to turn off soaring mode and make a dash through Montague (again, we obeyed all car speed limits!) and then had the speed set to the 45kt range for the last two miles or so. After logging the finish time, we still had some 1500ft and decided to scratch back to the airport. I made the airplane play Mr. Conservative again and we searched out some good lift to return the favor to the team that flew in the eagle thermal with us. Did I mention that we were only 2 minutes from second place? I fully believe if we hadn't spent so much needless time at the start of the course wallowing in sad lift, and if we had a better final glide calculator than "um, that height matches our L/D well, let's go now," I seriously think we could have found 2 minutes. Oh well, that's what next year is for!

So to conclude, the whole weekend of flying was a tremendous learning experience. The Montague Cross Country Challenge event brings a super crowd of people together in one of the prettiest places I've ever been. Dean Gradwell, who hosts the event, is a super nice guy and you can really tell he likes seeing folks having a good time flying cross country. I can't imagine a more fun way to spend a summer weekend :-)

Thanks Dean, thanks Chris and Adam, and thanks to all the other pilots & teammates that I met. It really was a great event and I am very happy to have been a part of Montague 2008. I have a huge list of how to improve my algorithms and no suggestion is going by without some investigation. When you attend an event with the best RC XC pilots in the world, you have to weigh their opinions and critiques heavily! I believe a humble soul and open ears go a long way for making improvements. Thanks guys, much appreciated.

Okay, I think that's my story. I have been replaying the flights on the computer and finding all sorts of more areas for improvement. Also, playing John Elias' and Jim Rolle's flights superimposed together has been pointing out the lift-producing areas. I haven't figured out what I'm supposed to learn from this data yet, but I'm sure there's a clue out there somewhere. The bald eagle knew it. Hopefully one day so will I.

Green Air,
Dan

You can now download some Montague Flights! You will need to download the free program SeeYou (direct download) to view these .igc files. If you know of another igc reader and would like to share, I would be happy to post it here. One of the features of SeeYou is the ability to plot simultaneously multiple flights. This allows us to compare manual pilots versus the machine.

Friday - Speed Task
• Rolle's Flight
• Elias' Flight
• Edwards' Flight

Saturday - Distance Task
• Elias' Flight
• Edwards' Flight

Sunday - Speed Task
• Edwards' Flight
  

As you should notice from the animations, sometimes we do funny switch-backs or seemingly circle in sinking air. Yup, I have more work to do. This algorithm is a work in progress, so it does include some quirks. The core part - how to actually use a thermal for gaining energy - you can see works rather well though. The mission-management for a competition scenario is a lot harder than I thought it would be.

Some of the lessons learned from the videos: our flight path is about the same sink rate as Elias' plane. That means we're fairly equivalent in L/D and glide path control. Sometimes ALOFT beats him since manually it is hard to fly a perfectly straight line. His speed-to-fly intuition is better at adapting to the local conditions quickly. ALOFT does much prettier circles, but Elias seems to find usable scratch lift down low a lot better. More thoughts........? Feel free to suggest!

Some of the future to-do items:
• Improvement on the engage/disengage logic
• Left/right turn direction determination
• Search pattern?
• Better mission management such as track to point rather than track to path
• Use flaps to change camber optimally!
• Speed gain tuning improvements
• Investigate mode
• Density altitude checks for high-altitude correction factors

I'll have more updates likely later this week.
Dan