AGU Data Dump

Disclaimer: This post is as much for my own benefit as anyone else's. I needed to write down all the things that are swirling around in my head after AGU and I thought it might be interesting enough to others to put up here.

First, AGU went really well. Here you can see the poster I presented.

I had the great opportunity to network with a lot of people working in the UAV field and there were quite a few people who had a great deal of interest in what we are doing. I was able to meed John Dandois of Ecosynth and the University of Maryland (ecosynth.org), one of the true pioneers of using UAS technology in ecology. They are doing really great work on the same sorts of things as us and I definitely see the possibility of some collaboration in the future. I also met people from University of Arizona, Oregon State University, University of Alaska-Fairbanks, Norway, and San Francisco State University who are all working with UAV and ultra high resolution imagery. These connections all have the possibility for learning quite a bit from the mistakes of others and for them to learn from ours. Additionally, I had the chance to meet with some vendors, which is part of what I wanted to cover here.

I spoke with guys from Google Earth Engine and Microsoft Research. Earth Engine is really cool in that you can run some pretty advanced data analysis in the cloud, which would take a lot of the load off of us when generating really complex products with vast amounts of data. Then, with the ability to tie Earth Engine into Maps Engine, where we could upload imagery and have it processed by Google into more manageable pieces, the prospects for a really powerful data management, visualization, analysis, and dissemination tool start to become clear. This is something that I will be working with over the next few days and I am pretty excited about the prospects. Hopefully a grant from Google which will provide us access to these tools will really open up our possibilities in the data visualization area.

Microsoft Research has an interesting GIS program called Layerscape which looks to be much more user-friendly than something like ArcGIS and, as you might expect, it comes with all the necessary bits and pieces to disseminate the maps once you have made them. It looked pretty cool but it remains to be seen whether it will be able to beat the Google offerings and be worthwhile.

I also spoke with a couple hyperspectral imager vendors and people who are operating hyperspectral units on UAS, which got me thinking about these kinds of things again. I found a couple companies which have really impressive offerings, namely Resonon (resonon.com). I heard from a researcher I was talking to that the Resonon systems are actually fairly inexpensive, which would be huge for us if true.

Well, thats all for now, hopefully I have even more exciting stuff to share next time.

Drones in the news

Well, I thought I would take a moment to stray from our normal pseudo-scientific ramblings to talk about a few cases of drones in the news recently. First, I'll present to you an op-ed from the New York Times which was written in response to Jeff Bezos' announcement of Amazon Prime Air, a possible future delivery service that you have probably heard of already. Take a moment to read the op-ed and then continue on to what I have to say about it.

The op-ed is nothing more than vacuous drivel coming from someone with no more understanding of drones than of quantum physics, string theory, or (apparently) logic. After confessing her fear of skies thick with drones (never mind that our skies are already thick with planes with 9.8 million manned flights crossing our skies in 2012), she shows the value of her hand almost as soon as she starts, "So if they can’t land on my head, why do they make my head hurt? Maybe because they are redolent of President Obama’s unhealthy attachment to lethal drones..." I suppose she can't be bothered with the fact that the aircraft in the sky today have the military to thank for much of their development (First jet aircraft? Luftwaffe). To reject the use of drones based on their military application is tantamount to rejecting air travel because airplanes can be (and frequently are) used to drop bombs on people. In fact, if she could trouble herself to look at the numbers, she would see that of all the weapons released by aircraft in Afghanistan from 2009-2012 (18,559), only 6.8% of them were released by drones. How is it that they are public enemy No. 1, the evil machines raining death on the Middle East when they account for 7% of the lethal air power used and the other 93% gets a free pass?
After she finishes with her journalistic masterpiece about the potential lethality of these innocuous little aircraft she makes an entirely unrelated and unsupported statement about the "unhealthy attachment to indiscriminate surveillance" of our spy agencies. This is a matter of opinion, and many people agree, but it is quite out of context and not entirely applicable here. Not even the Snowden leaks could produce evidence of surveillance on Americans in America (by drone or otherwise), which is the only type of surveillance restricted by our laws. Our spy agencies act within our laws, so if you want them to act differently you might consider changing the law. I will readily concede that the idea that spy agencies may be tempted to spy on us from above is not unfounded but neither is the idea that drones could revolutionize our daily lives in ways that we cannot yet predict. The Wright brothers gave rise to next day air freight, intercontinental air travel and, yes, even drones. Do you think they would have expected that?
And, as if her murderous privacy-invading drone point wasn't clear enough (It was plenty clear, just not logical) she goes on to provide some tenuous link to the latest episode of Homeland. I won't even bother addressing the absurdity of connecting reality to a TV drama.
After all of that nonsense she does manage to put together part of a factual and sensible statement. She says that Bezos is simply putting on a PR show and catering to our desire for instant gratification. This may be true, but that is his job; he has built his entire company on the foundation of providing anything you want as soon as you want it. The highly successful Amazon Prime is centered around providing anything Amazon sells in 2 days at no additional charge. Can you blame the man for pushing the delivery time envelope? Further, while the items she might order would be argyle sweaters, some of us (like us in the lab) order things that we can't get locally and need yesterday. Turning 2 day shipping into 30 minute shipping has the potential to turn 2 wasted days into barely enough downtime to eat lunch. And her implication after this that other companies following suit is also a bad thing is equally off base; why is a coffee being delivered by drone any worse than a sandwich or a pizza being delivered by bike or car?
She then goes on to mention lobbying and movie producers and some stuff about jobs. To be quite honest I'm not sure what she was doing here. She provides some of the biggest reasons drones should be integrated into the airspace (100,000 jobs, $90 billion industry) with no explanation, no support or rejection. Then, with a similar nonsensical transition, she returns to Amazon's drone delivery and a quote from Politico (a political news outlet with presumably as much drone expertise as our dear friend Maureen or Barney the Dinosaur) that highlights the danger of operating drones in a city. Sure, current generation drones probably aren't reliable enough to safely conduct this sort of task, but does that mean the drone of 5 years from now will still be incapable? Does our current track record of dizzying rates of technological advancement suggest that we will make no meaningful progress in this area in the next few years?
Next, she is on to another "drones will take away your last shred of privacy" argument. I would refer her back to the second paragraph for a refresher on why this isn't a sound logical argument.
Finally, she goes back to breaking down her own case with a few of the multitude of valuable uses of civilian drones before making another absurd reference to Hollywood as if it has any relevance to reality. Her op-ed ends just as weakly as it started; same old fearmongering tactics, new half-baked examples. I understand that people are afraid of change but have we not seen for a hundred years or so that our fears are generally alleviated by overwhelming benefit?

Phew, ok, now that I have that out of the way, I'd like to turn to some brighter news. First, Lian Pin Koh, the director of ConservationDrones.org, recently gave a talk at TED in Edinburgh, Scotland. His talk is certainly inspiring and shows the real power of the aircraft we work with, but I was a little disappointed that I wasn't invited to speak. I'm kidding, kinda. While the work they are doing is impressive it in no way exceeds our accomplishments and I would even say that we are quite a few steps ahead of them, as we are much more advanced in the area of multirotor-based imaging. They have gotten their aircraft out to more places and collected more data, which definitley counts for something, but the systems they have developed do nothing that ours can't. This is actually heartening for me, as it shows that the progress we have made over the past 16 months is really meaningful and basically TED-worthy.

Finally, I'll share a few stories of us in the news. We have already gotten some good attention as a result of these and I really think it will help us as well as drone development in general.

WFU News

Science Daily

El Comercio (Peruvian equivalent of the New York Times)

Old Gold and Black Newspaper

Consumption Resumption

I have to admit I just sat here for about 5 minutes trying to think of a good word that rhymed with presumption and assumption. Anyway, as the title might suggest, I'm bringing back our earlier discussion of power consumption. When I left off last time I noted that slowing down by about 30% can more than double our flight range. This got me thinking, what other small tweaks can we make to get big gains; the low-hanging fruit, if you will.

I started to think about motors and propellers and their efficiency. With that in mind, I pulled up eCalc, a great online tool for calculating all sorts of things about your aircraft, motor, propeller, and ESC. The big thing I was interested in was efficiency of the motor/propeller combo. I set out to model what we currently fly, an MK3638 motor with an 11x6 propeller. Here is the graph I got:

As you can see, the efficiency of this motor/propeller combo (the blue line) peaks at about 90%. It's 87.8% to be exact. It also has a nice thrust-to-weight ratio of 0.66:1. Those number are both quite good. The only problem is that the 87.8% efficiency doesn't happen until you draw 24.1 A. This is a problem because we generally cruise at 5-10 A. As you can see in the figure, the efficiency in that range is pretty bad in that range (well lets just face it, for an electric system its abysmal). At 10 A the efficiency is about over 77% and in the butter-zone, 7.5 A, it drops to less than 75%.

So, what is the diminution solution? Enter, the Tiger MT-2814-10. This motor spins at 770 RPM/V, just like the MK3638, but the way it is wound does something interesting with it's efficiency curve:

Again, the peak efficiency of this setup is quite good, 86.2%. It also has a decent thrust-to-weight ratio of 0.60:1 (with a change to a 12x6 propeller). The icing, though, is that it's 86.2% efficiency peak falls right in the middle of our cruise range, 8.8 A. This means that with a simple motor/propeller change we should gain more than 10% efficiency at cruise. For such a simple change that is a pretty impressive gain.

So, we ordered the motor and a propeller and for $60 we have (hypothetically) added 10% efficiency. I got it mounted up yesterday and spend a bit of time today balancing the propeller and motor. It is now ready to fly and, as an added benefit, it puts almost no vibration into the airframe (look here for a bench test after mounting and balancing which shows that there is no longer any "jello" in the video). Now, all we need to do is get it airborne and see how it goes.

"It just won't flat spin"

Yesterday was an outstanding day. We joined some other local FPV flyers for a small fly-in at the RAMS RC club (the field where we normally test our planes). We took the DAO, the Phoenix, and my plane (an FPV Raptor), which I'll just call the Raptor for now until another name presents itself. We were also joined by Kevin Wang, the newest member of the lab, a freshman who is interested in computer science and wants to work with us to enhance our systems. We set out to do a few things:

1. Get me some flight time on the Raptor, both LOS and FPV
2. Return the Phoenix to the skies
3. Do a long range flight with the DAO, simulating a flight to Pilot and back.

On all three counts we succeeded. From this point on I'll tell the story with pictures.

We arrived and promptly set up the "Christmas Tree Farm," our array of tripod mounted antennas and video monitors. As you can see here, the skies were pretty clear and the wind was light from the North (from the direction of Pilot).

Kevin got to learn from the best.

The Phoenix was revived and ready for action.

I wasted no time getting the Raptor ready to fly.

The Raptor took to the skies after an FPV takeoff and gave us great FPV views of Pilot under the wispy clouds. The flight was a completely uneventful (a good thing) 40 minutes of FPV bliss. After I had my fill of flying I brought her in for a landing under FPV.

The DAO, unmodified since our 7.25 km flight, was sent skyward with Tim at the controls.

Our flight plan was to travel out to the west of Pilot, pass it, and return. Pilot is the large dark green forested area at the top, west of Pinnacle. As you can see, we returned a little short (about 2 km) of our target. At about 12.6 km we started to get some spotty reception on our Dragonlink. We made it out 12.8 km from home. Surprisingly, the FPV system (800 mW 1.3 GHz tx, RMRC rc w/ SAW upgrade, homebrew half-wave dipole on tx, and homebrew biquad on rx) gave us totally rock solid link for the entire journey (except for turns when the bank angle turned the tx antenna out of polarization). The RFD900 telemetry system and yagi also gave us perfect performance, with Mission Planner reporting 98% RSSI the whole time. We plan to remedy our Dragonlink reception issues by flying higher and using the Moxon (more directional) antenna.

If you happened to be on the observation deck at Pilot, this is what our flight path would have looked like.

Here you see Pilot from the DAO shortly after takeoff. Range to Pilot here is ~13.5 km.

Pilot mid-flight, just about to pass over the bend in the Yadkin River and leave it behind for bigger, better things. Range to Pilot here is ~9.75 km.

Pilot again mid-flight, passing over some farmland. Range to Pilot here is ~7.75 km.

Pilot shortly before turning back for home. Range to Pilot is ~3.0 km. Range to home is ~12.5 km.

And she returned home safely. The RAMS is the green grassy area just below the center in this picture.

After flying the DAO, it was time to see what the Phoenix could do. Here you can see it leaving Tim's hand on its re-maiden. Unfortunately, the settings in the transmitter had not saved after Marcus changed them and the ailerons were reversed. This led to a very quick but spectacular return to earth with minimal damage.

After some quick repairs to the tail and the necessary adjustments to the transmitter, the Phoenix was sent skyward again. She flew like a bat out of hell. She has tons of power, can almost hover, knife edges well, but doesn't like to fly inverted. Also, according to Marcus, "It just won't flat spin." I should go ahead and note here that we were not, in fact, having fun but were testing the full range of flight capabilities of the aircraft. The big news here is that we have another functional aircraft which will be useful for testing new features, firmware, techniques, etc so we don't have to use the DAO for that anymore.

Here are the DAO and Phoenix after flying, in good shape and ready to fly again.

Stay tuned for video from the DAO long range flight and the Raptor.

Antennamania

As promised in my last post, here is a small write up on my long night of antenna building last night. Partly inspired by a perceived need to boost our FPV range and partly inspired by an ever-growing list of things I really didn't want to do, I decided to build some antennas. This was my first foray into antenna building and it all still seems to be black magic to me, so I stuck with the very well put together tutorials by IBCrazy (Alex Greve, pretty much the undisputed leader in FPV antenna development). Over the past few weeks and months I have thought a lot about antennas and what we want out of ours, so when I decided to build some, I had a decent idea of what I wanted. I wanted to make a highly directional 1.3 GHz FPV antenna for our receiver, a 1.3 GHz FPV transmitter antenna with good long range strength, and a more directional antenna for our UHF (433 MHz) RC transmitter. Below are my antenna choices.

For the directional receiver antenna I decided on the biquad. To me it was appealing because it is simple, elegant, fairly easy to build, and has 11 dBi gain (effectively doubling the range of the 8 dBic patch antenna). Below is a picture of the antenna; find the link to IBCrazy's tutorial a the bottom.

1.3 GHz Biquad antenna. 11 dBi gain. Using RG58 coax, 10 ga solid copper wire, and steel sheet.

For the video transmitter antenna I decided on a simple half wave dipole. My reasons were simple, the antenna has a toroidal radiation pattern which puts its radiation out in a 360 degree beam but not vertically up or down (which would be wasted energy for us). It is also exceedingly simple to build.

1.3 GHz half wave dipole. 2.15 dBi gain. Using RG-316 coax, 18 ga stranded wire, and a protoboard.

For the UHF transmitter antenna I decided on a Moxon. This is basically a Yagi with no parasitic elements (only a driven element and a reflector). It is ideal because it is, again, pretty easy to build and significantly more directional than the stock antenna. It has a gain of 5.75 dBi. This one is still a work in progress, as I decided to prioritize the FPV antennas. I'll get picture of it up here as soon as its done.

When it was all said and done, this is what the plane looked like. The GoPro is mounted using 2 velcro straps passed through holes cut in the FPV plate. The dipole was mounted on the left side of the nose (right from this point of view) just below the GoPro.

The transmitter and battery were mounted to the bottom of the FPV panel using velcro. This made for a very compact package which can be readily removed from the aircraft.

Video transmitter antenna tutorial (includes 2 types of half wave dipole and the inverted vee)
Biquad video receiver antenna tutorial
Moxon antenna tutorial (includes 3 and 4 element Yagis as well)

UPDATE: The BiQuad is an amazing antenna. It, along with an RMRC 1.3 GHz rx, RMRC 800mW 1.3 GHz tx, vertical half-wave dipole, high pass filter on the rx, and low pass filter on the tx gave us 12.8 km of absolutely rock-solid video. We were flying in clear conditions 250 m AGL with the rx antenna 2 m AGL. We lost Dragonlink 433 MHz connection and were forced to turn around before getting so much as a hint of degradation on the video.

Consumption Presumption

After spending the morning at the field with the DAO, I have multiple items of pretty good news to share. First, FPV with my new DIY biquad antenna (more on that in a later post) worked flawlessly on the plane. Second, we found flying with 12,000 mAh of 4S to be a piece of cake, we didn't notice any ill effects and, as you'll see below, it promises quite a bit of additional flight time. Finally, it seems as though we have remedied the problem of high power consumption on the new airframe. It looks as though the issue was pretty much unrelated to the airframe and instead a function of speed. When we flew on Tuesday we were cruising at ~14 m/s. By backing off the throttle a little today and keeping our cruise speed at ~10 m/s, we greatly extended our flight time. With those pesky laws of physics increasing drag force by the square of velocity, we really hurt ourselves with those 4 m/s. Check out below for a graphical representation of today's results.

This figure shows battery consumption during our first flight. The initial constant-slope portion is battery being burned sitting on the ground. From this we can see that its pretty important not to do that if we want peak performance. The rest of the slope is remarkably constant, despite the fact that this flight included a climb from 100 to 350 m over 2 km. That particular climb can be seen as the inflection point about 3/4 of the way through the time interval. You can clearly see the downward turn as it climbed and the flattening out as it rapidly descended.

This figure shows battery consumption during our second flight. During this flight we upgraded from 4S1P to 4S2P, for 12,000 mAh of capacity. As you can see, the slope is quite constant , but slight variations in slope can be seen as we climbed from 100 m to 200 m, descended, flew level, and then repeated the process. Also, don't pay much attention to the appearance of a large difference in slope between the two figures; they are shown over significantly different time intervals.

Our first flight, with lengthy loitering around the field, and 2 passes of a flight plan. The first pass took us from home to the right at 100 m, turned up river and climbed to 350 m, and then turned back for home while descending back to 100 m. The second flight was the same but the climb was reduced to 200 m. The flight path from the south to north end is 2 km. Total flight time was 30 minutes and the total distance traveled was 20 km.

Our second flight, with 4S2P configuration. The flight plan was the same as the other, with a climb to 200 m. Total flight time was just under 20 minutes and the total distance traveled was 14.5 km.

So, with all of that said, here is the really good news. With the 4S2P configuration we should be able to fly at least 45 km. We expended 26% of our battery over 15 km, which means that 45 km would leave ~25% in the tank (5% reserve). This calculation would also seem to be fairly conservative, as a climb to 200 m followed by pure straight and level flight should be significantly more efficient than climbing, descending, and circling. So, having that number, Miles and I did some back-of-the-envelope calculations and, based on that flight distance and the footprint of our images (with overlapping areas removed) and came up with 450 as the number of hectares we can map in a single flight. If we map 450 ha in a 1 hour flight and fly 4 times in a day, we can map 18 square kilometers in a day. I don't know about you, but by my standards that is a pretty big area.

Consumption Assumption

Here is the remaining battery capacity vs. time for the previous 7 km out and back..  I have marked the start, turn around and end times.  

The rate (slope) consumption is fairly constant as a function of time in the upwind and downwind directions. If air speed is governed to ~12 m/s power consumption should be fairly constant.  If you were to plot remaining battery vs. distance traveled then you calculate:


For 6000 mAhr
upwind rate = 42%/7km = 6%/km
downwind = 26%/7km = 3.7%/km

So using 14km to turn around:  6%/km * 14km = 84% would be used of 6000 mAhr to get there
and 3.7% /km * 14km = 52% of 6000 mAhr to return.

Which looks possible with 12,000 mAhr and no climbing.

We could refine this and have a two part equation by testing consumption in a sustained climb with 12000 mAhr of batteries.  Based on the above you have about 30% of 6000 mAhr to climb about 900 m over 14 km.  Climbing into the wind will definitely help so we are on the edge of plausible.  I would like to include wind speed into this estimation at some point.  

Making some assumptions here but that's better than just hoping.

Flight paths from yesterday

As promised, here are images of the flight paths from yesterday.

Out-and-back flight reaching 7.25 km linear distance from home. Pilot Mountain is marked in the upper-right corner. Colors indicate flight modes; green is the auto flight out, purple at the end is a small guided-mode addition to increase flight distance, and pink is return to launch. County lines are denoted in light blue so, as you can see, we just barely made it into 3 counties in one flight. This flight was preceded by a 4.5 km out-and-back flight, not shown.

The latter portion of our copter flying, showing takeoff, a short flight out with return to launch (red), followed by entrance into the autonomous flight of a grid.

Big News

Well, suffice it to say a lot has changed since my last post. We made a run at Pilot and, due to a disconnected airspeed sensor, ended up in a full power descent into a stand of trees. The Catfish has been rebuilt, and has a new name, the Phoenix. It hasn't flown again yet, but we are working on that.

We took the opportunity provided by the crash to get the newest version of Skywalker frame, which has a much more spacious fuselage. The space makes it easier to work with and also provides us with the room to carry dual 4S 6000 mAh batteries. As shown a few posts ago, we had our maiden flight last week and got it flying well. We did have to compensate for some warped wings with a lot of aileron so I worked to get that removed during the last week. Today we sent it on its first Auto flights since the crash.

Due to continued frustrations with the MikroKopter system and astounding success with APM on our hexa, we made the decision to switch the APM over to the octocopter and see how it performed. We had immediate success in stabilize and loiter, with very little modification of the hexacopter parameters required. Today we set out to test RTL and Auto.

Finally, after coming into a small sum of money (no sketchiness was involved), I sprang for an FPV transmitter. receiver combo and the cables to hook it up to my GoPro. I also integrated the EagleTree OSD that we had laying around the lab collecting dust. I got the system assembled and installed in my plane over the past few days and flew it for the first time today.

Also, to cap the day off, an article about our work went live on the WFU homepage today to complement another article that was in the Old Gold and Black last week.

Read on for more details about our day today and the culmination of these efforts.

John 11:43 “Lazarus, come out!”

After 4 hot glue sticks, half a bottle of epoxy and small bottle of CA.  Catfish is on its second life.  Will need to re-maiden and see how it flies.

New 2013-SW madien

To do on New Airframe:

1) Horizontal stabilizer not parallel with wing. 

2) A lot of left Aileron and some left rudder trim needed.  Looks like the right wing has more wash-out (right front tip is twisted down) causing the right roll.

3) Some right roll seen in FWBA and Stabilize.  Running level calibration seems to have helped.

4) Elevator push rod is flexing with push.  Also need to reset linkage for the servor's neutral point.

5) Geo fence not responding to switch.

6) Glide ratio is less, not sure if it's due to drag from trims.  Better land with power on.

7) Much more motor authority with 12 x 6 APC.

In spite of these mostly minor issues we flew this Auto mission in light wind (5-10 mph).

Round and Round

APM Hexa did a nice circle WP filight in 10-15 mph winds.  Not bad.  We have come a long way in our team effort approach.  Max building, Tim tweaking and me not wanting to give up.

Next, we did three our four auto take off and landings.  It was such a rush:

These are the points where it went up and then landed in auto.  

So now we need to fly, fly, fly.  Put a gimbal on it and fly some more.  We need to push it to failure.

Exciting news and a big day ahead

Well, I have a few exciting updates to pass along as well as our plan for a big day today. This past week we made a lot of progress on multiple projects.

• The frame for Dave Anderson and Felipe Estela's hexacopter arrived and Felipe and I got it assembled. It is an all carbon fiber Tarot 680 folding frame and seems to be of pretty high quality. I will give more updates as the rest of the parts arrive.
• Altitude hold on the Ardu hexa has been sorted out and now we are working on loiter and RTL. We had a successful RTL but some odd behavior on our second attempt. Definitely stay tuned for more updates on this as we continue to refine.
• We got a new navigation controller for DeaconEye and that seems to have remedied our compass issues. We also learned a lot about compass interference and our magnet errors at the same time. We will be working towards returning DeconEye to dully autonomous flight this week.

Now, for this afternoon's agenda. We are heading out to the RAMS field in about an hour to fly the catfish to Pilot Mountain. This is a 13.5 km trip each way and requires us to climb 500 m from our normal cruise altitude. It should be an excellent test of all of our long-range systems. Come back this afternoon for updates on this flight.

Flight Logging

So, lately we have been doing a pretty good job of keeping the blog updated with the big news coming out of the lab. Something we haven't been doing so well with, however, is routine logging of our flights. In the beginning, we did a great job of this but we have tailed off recently (mostly since Dave left us for the nether-regions of New Hampshire). So, in an effort to streamline the process and get us keeping up with it, I have created a new form on Google Docs to make it easy to do and easy to pull data out. The form tracks an array of data points, including flight times, what aircraft are flown, weather conditions, batteries used and their charge condition before and after flight, and, of course, extensive error/failure tracking. All of this data populates into a spreadsheet and an easy-to-use summary page, which will show us things like cumulative flight times and rates of in-flight incidents. And the best part is that it is all accessible from an iPad in the field. So, with all of this in place, I hope that we will have a safer and more efficient operation, and we will be able to readily prove that we do.

If you're interested, take a look at the log system below:
WFU UAS Flight Logging System

DeaconEye Lives!

So, after a long wait with a lot of hours of labor and a significant amount of cursing (p=0.00001), DeconEye has returned to the skies. Today I put together a new power distribution system, which I am dubbing the octopus (although it really has 18 arms, 8 power and ground to each motor along with an accessory power lead), and got everything ready to fly. I calibrated all ESCs, changed the necessary settings in MK Tool to get it playing nicely with the new I2C-PWM converter, made sure the motors were all spinning the right way, and got the props balanced on the motors. I also checked to see if the problems with the compass heading varying wildly had been resolved by replacing the old ESCs, and it indeed was. You can see a picture from immediately before the flight below.

With all the details attended to, I took her out for a flight. As soon as I got in the air I could tell we had made the right choice. Everything was superb, it flew as expected at all time, was extremely stable, handled rapid descent very well (I think better than with the old ESCs), and behaved excellently in position hold and altitude hold. At this time I see no need for any additional tuning to get it flying well, it is in excellent flying condition. Return to level when the sticks are released is a little sluggish, but I'm not sure it wasn't that way before and the snappy hexacopter just changed my expectations. My laptop was acting up, so I wasn't able to do any waypoint flight, but from what I saw in position hold, carefree, and altitude hold, I have no reason to expect any problems.

There are, however, 2 significant differences from the old system which I will note: the motor idle speed had to be raised and is double what it was before, and the aircraft now hovers at slightly over half-throttle (55-60%). The first issue comes with the new ESCs but I will work on the second to get the hover back at half throttle (probably by doing a simple travel adjust on the transmitter). I plan to spend next week getting a significant amount of flight time to make sure everything is as good as it seems and then it will be time to get some footage of fall colors on campus.

Objectives

Well, it is time to lay out some objectives again as we need to keep our focus and continue moving in a productive direction. In this post I'll provide a brief update on the state of the lab and outline some short and longer-term objectives.

The Catfish is flying well these days, and all testing has pointed to good, highly reliable flight. Marcus constructed a new pan/tilt gimbal for the nose of the plane to hold the camera in a position that will allow us to image the knob of Pilot Mountain. It was test-flown on Saturday with excellent results. Everything else with the Catfish is squared away, except for camera triggering. We lost that function when adding the gimbal, as we had to move the servo connections around. The good news is that it should be trivial to regain this functionality, it will just take a few minutes of doing. The use of the airspeed sensor and Total Energy Conservation System (TECS) tuning is on hold until we have flown Pilot, as we want to fly at the peak of fall colors and can't afford to break anything now.

DeaconEye is still grounded, as I fried a power distribution board on Friday and we are working on putting together an alternative solution. Everything else on it should be ready to fly, and once we get the power situation worked out it will be time to re-tune the aircraft with the new ESCs.

As shown in the posts below, we have imagery from Duke Forest and analysis of this is pressing. From this imagery and the imagery collected with DeaconEye, we want to start producing real data. This is what much of the "off-season" will center around.

Finally, the deadline for the National Geographic grant is almost here (October 22nd). Our pre-proposal was approved and now we need to put on the polish and get it submitted. Simply put, we can't let this one get away.

So with that recap, here are the objectives, in a rough order of importance:

Near-term (next 3 weeks):

1. Develop and implement a power distribution solution for DeaconEye
2. Re-tune DeaconEye
3. Finish National Geographic grant
4. Fly Pilot Knob

Long-term (next 3 months):

1. Delineate tree crowns in Duke Forest imagery and correlate data to known tree data
2. Apply canopy illumination model to 3D canopy model
3. Implement Catfish airspeed sensor and TECS
4. Fly somewhere exotic (Peru?)
5. Procure fluorescence sensor
6. Procure 20 mm prime lens for Sony NEX-5R
7. Procure low-light/night vision/thermal recording and FPV system

Pacific Winds are Favorable

One of the missions we want to fly is out over areas in the middle of the eastern Pacific.  One of the problems with flights in oceanic areas is that the winds can be strong, and it is hard to get information for a lot of the earth on the scales we need it.

I was poking around the National Data Buoy Center for windspeeds this morning.  There are buoys tethered out in the ocean that measure climate and oceanic variables for studying ENSO.



Here are the daily mean wind speeds from April-today measured at 4m for buoys at:

8N, 95W   4.3 m/s  9.6 mi/hr  <--- west of Cocos Island
2S, 95W   5.0 m/s  11.2 mi hr <-- west of S. Galapagos
[Continued below the fold]

DeaconEye Overhaul

DeaconEye mid-overhaul, shown with the new ESCs, new power distribution board, and a motor ready for new bearings.

Well, what started as a handful of upgrades and repairs has turned into an almost complete overhaul of Deaconeye. Here is an outline of what has been done and what is in progress (Continued below the fold):

Duke Forest

Well, it has been a little while since the last time I posted and there is quite a bit to talk about since then. First of all, Marcus and I went to Duke Forest on Saturday and flew all of 4 of the plots that we wanted to fly. In short, it was a pretty incredible day of flying with ~70 km of linear flight in ~120 minutes. We imaged 2 plots per flight and used the "multilooked" method Miles discussed earlier on the blog. Here is a look at our first flight:

You can distinctly see where we took off in the open field as well as the two plots we imaged in this flight. The colors indicate flight modes, with orange being manual and the green being auto. Below is the profile of our second flight:

The images from these flights are looking pretty good, although we have a clear need for a higher-quality camera. Below you can see the preliminary mosaic of our second flight:

This is only a subset of the mosaic which excludes peripheral areas which are not in the area of interest and are not imaged well. You can see a few errors along the road at the top; Photoscan is known to produce errors along straight lines.

Finally, here is what I am affectionately calling the Asnerian View (an oblique view of the 3D model):

So, from this we have learned quite a bit. We have learned that the camera we are using in the Catfish has a few idiosyncrasies which we are learning to overcome along with a diminutive sensor which makes collecting good imagery on anything other that a bright, sunny day quite difficult (the sky when we made these flights was overcast). All in all, though, we have proven that the Catfish is a really great, reliable system for imaging large areas in a short amount of time. Our next project for it will be to go to Pilot Mountain and image the knob.

Upgrade Time

We have taken the downtime resulting from our close encounter with a tree with DeaconEye and the wait for an airspeed sensor for the Catfish to work on some upgrades for both.

The Grind

Today we're hard at work on the daily grind. Hand-balancing propellers, installing a new gimbal motor on DeaconEye, and repairing a faulty connection in the Catfish. The work in the SUAS lab isn't always glamorous, but it always brings us closer to our goal of safe, reliable flight collecting usable imagery that helps us answer the research questions. Who knows what ecological breakthrough will come from the balanced prop?

Good news/bad news

Well, I'll start with the good news. Today we took Catfish out to the RAMS field to dial in the APM settings and had some outstanding flights. We went through the APM recommended tuning procedure, tuning roll, pitch, yaw, and autopilot navigation. The results were so good compared to the out-of-the-box settings that Marcus even said he might like flying in fly-by-wire mode better than manual. We got response in FBW to be crisp and manageable, something we have not seen before. We got flight in auto mode to be extremely accurate in most cases, making tight turns to stay on track and hugging the planned flight lines. We also

New Video

Check out this new video I just finished cutting together. It is a compilation of footage from 3 days of flying. The video is straight from the camera and cut together, there is no post-processing vibration filtering here.

A Great Day

Today was, in short, a great day. As a recap, we have have been having some issues with our aircraft lately. The majority of the Catfish's avionics (APM, RC receiver, telemetry module, 3 servo motors, GPS receiver, ESC; so basically everything except for the motor and one servo) were destroyed when an exposed pin on the RC receiver (a ground pin) punctured the silicone insulation on one of the main battery wires (the hot one) while the battery was being put into the aircraft. This resulted in some fireworks and lengthy repairs. We got the Catfish back in the air a couple weeks ago and it flew great, especially with the new metal gear servos and new ESC. The plane has much more control authority now compared to when it had the cheap, very low quality servos.

A good day at Reynolda Meadow

Today was a good day for all aspects of the DeaconEye project. We had 2 missions for the day, to collect imagery of all of Reynolda meadow for mosaicing and to do a thermal imagery trial run to assess the ability of DeaconEye to collect good thermal imagery. The day really started yesterday, with us making 4 test flights (~1 hour flight time) to ensure that everything was in good working order and we would have no surprises today. We worked out the last of the kinks related to the issues we had with magnet errors and poor telemetry connectivity and ended yesterday feeling confident about the state of everything.

Today we started out by collecting the RGB imagery of the meadow. We imaged the entire meadow in 3 flights (if you don't count one aborted flight, more on that later) using 3 sets of batteries. We made the first 2 flights on the Thunderpower 6600 pair and made the next 2 on Hyperion 5000 pairs. We elected to change batteries after we aborted a flight a couple minutes in to ensure that we did not push the envelope.

Big Deac

Big Deac is up, running, and fully operational. System specifications are as follows:

• Dual Intel Xeon E5-2620 CPUs (6 cores each at 2.0 GHz, 24 total virtual cores)
• 64 GB RDIMM ECC DDR3 RAM
• AMD Radeon HD 7970 GPU (32 OpenCL cores, 3 GB DDR3 RAM)
• SuperMicro X9-DAI motherboard (4 PCI-E x16 slots, 16 DIMM slots, 10 SATA ports)
• 2X 2TB Western Digital Black HDDs (RAID 1, for long-term storage)
• 128 GB Samsung 840 Pro Series SSD (Soon to be 2 in RAID 0, for working with data)

As the specs above might betray, Big Deac is a beast and has laughed at most things we have thrown at it so far. While using Photoscan we have now found that we are software limited and no longer hardware limited. We are capable of processing image sets of up to 1000 uncompressed TIFFs on the highest quality settings,

No good turns and multiple looks

One of the things we have been working on this summer is bringing online the fixed wing sUAS, variously known as The Catfish, Mr. Chow, Local Hawk, or simply "the fixie".  It is a foray into more travelled ground for sUAS in general than DeaconEye, but it brings with it a new set of problems for image analysis.  New cameras to be triggered, and new ways that imagery varies among missions to deal with.  One of the things that is very clearly different is that the fixie doesn't "mow the lawn" like a copter.  Turns put the un-gimbaled camera out of angle enough that we're not mowing, we're farming.  And when you're sowing, your turns at the end of the field are out of bounds--you use them to get the machinery turned around for the straight rows--and then you come back and plant the edges.  That's what we need to do with Local Hawk.  Set the straight parts of the flight plan to capture the imagery, and make the turns out of the desired scenes.

More generally, it'd be nice to start a set of python scripts for our image workflow that looks at the EXIF information and automatically leaves out pictures beyond a predefined bounding box for the imagery, and, given some of the severe cross-tracking in high  winds and roll during gusts, takes out images with high yaw

Making the imagery accessible

So in the time that we've been offline we've increased our stable to include ArduCopter and ArduPlane, and climbed the learning curve on the minutiae of MikroKopter to the point where we have it flying autonomously to take high resolution canopy mosaics and generating DEMs of canopy surfaces.  All very cool.

But this generates enormous amounts of data, and we have to be able to process it  and display it efficiently. Not only in ways that are good for us in terms of the science goals, but also in ways that people can use on the web.

So how to do it?

One of the ways that we are all familiar with that uses tremendous images in a friendly way is through Google Maps and Google Earth.  And the Google Map API can be used to serve and navigate our own images.  Let's take a large mosaic--in this case a section of Long Caye in Belize--and

4WD Mowing

I knew it would be an interesting day of flying when you have to put it in 4WD to get to the flying site.  Four huge Poplars across the road diverted us out to the Yadkin river and back.

After a Deans soldering marathon and Max putting  in a lot of work to ready the copter we loading up and moved out.  Max had us mowing the lawn again.  We imaged about 7 ac in no time as a test.  We learned the new 12v BEC interferes with the compass and we need to either move or shield it.  We also learned the copter uses about 500 mAhr/min.  So every 1 Ahr gives us about 2 min of flight.  Oh and MAXOsnapper worked perfectly. 

The Arduplane, now called "Catfish"  was a little terrifying.  It was not set up for the wind and it struggled to make way points.  We almost lost it the trees but Max saved our bacon and yelled "Put it in manual".  We have some work to do on speeding the camera up as well.  Max got some stick time on my little home built Tricopter and the Arduplane.  He had a real nice save on th plane.