New videos available! Check out http://www.bu.edu/im...ssful/#more-123 for some really cool FLIR action, including a crash caught in 135fps.
Braving the Freedom Grope and the other inane US airport security, last week Monday I walked into Boston Logan with a bewildering array of quadcopter parts, spare parts, bird netting, and a 12-pack of unlubricated condoms (for another bat experiment, unrelated to the UAVs. Promise.), headed off for Austin, TX. The objective of my week in the Texan back-country was to perform some experiments on trajectory planning in bats, alongside some other bat researchers from Boston University.
These Brazilian Free-tailed bats (also called Tadarida) come together in the millions in caves all over Texas, leaving every night in swarms so big they can be detected by doppler radar. Somehow, they manage to fly through this dense self-clutter without major collisions, and so our goal is to better understand this behavior. The goal is to fly a UAV through the dense clutter, and record the bats' response with three ground-based high-speed FLIR cameras, and an airborne 3D HD GoPro camera. The hope is to extract fundamental control laws of flying behavior in order to achieve better autonomous UAV flight.
P5200067.JPG 615.29KB 125 downloads
My marvelously talented team of lab assistants and I initially built an aluminum framed quadcopter, with the frame arms sourced from Home Depot chromed aluminum towel racks. The idea was to be able to find any part that we might break in the local hardware store. The netting was held in place through a box made of fiberglass kite rods, held to the quadcopter frame by bamboo rods. (Bamboo is amazingly strong, light, and can be found at any local garden store.). The overall weight of the frame was some ~600 grams (heavy, but that's what you get when you build out of parts available at the local hardware store) and the netting frame was ~400 grams. Add in the ~300g for the 3D GoPro camera (not the best HD camera, but what other 3D camera could have survived the abuse we put it through?), and you've got a lot of weight for just four motors.
While this design performed acceptably in Boston, the difference in temperature and altitude drove the AXI motors over their limits. Even with larger 12x4.5 props, the Batcopter v1.0 could not maintain orientation in the strong winds. In my opinion, this was due to, quite frankly, the fact that the CopterControl is not performing model-specific control. There was very little room for error, and the lack of coupling between motor and controller made it impossible to react sufficiently quickly. This is definitely something that the OP-ESC will improve significantly.
After several days of trials and errors, I finally gave up on the aluminum design and got to thinking about what I could do with the four 1m long carbon fiber rods I had brought down with me. I decided that the best approach was to build a V-frame that could eventually be upgraded to a hexacopter, with hopes of finding a sixth motor (I'd already brought down a fifth motor as a replacement in case one of the QC's motors was broken in a crash).
However, as I was thinking about the best way to join together my 10mm rods with my 5mm rods, I had a sudden flash of inspiration for how to do the netting. Along with all the other tools and supplies I had brought through airport security, I'd brought two pieces of blue foam that I'd be using as landing gear. I realized that these pieces of foam could probably serve to hold the netting far away from the props. This would mean that I could completely get rid of the fiberglass structure.
The next morning, I drove the 50km into town to raid the local hardware store, the local WalMart, and the local Radio Shack (hey, even in Middleofnowhere, Texas, they've got a Wally-world!) for supplies. At my first stop, the Ace Hardware store, I chanced across a real guy who actually had a clue-- lesson: shop less at faceless Home Depot and more at knowledgeable Ace. He and I brainstormed for a while, and wandered around looking for stuff that could work to attach the carbon fiber rods together. We decided that the best strategy was to use twine impregnated with white glue. The twine would be strong and the white glue could be dissolved in case I wasn't happy with the approach. It was light, strong, somewhat rigid, and would build up the joint so a heavy-duty zip-tie could be added for additional clamping power.
Having had a very successful, if long, visit to the hardware store, we headed off to the next stop, WalMart. However, along the way there I saw a very traditional sight in Texas: a gun store. And written underneath GUNS!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! was arrows. (Okay, I exaggerated, in reality there were only 46 exclamation marks, not 47.) Inspiration flashed again, and I remembered from my youth how the only model airplane that ever withstood the shear number of crashes I put it through was a flying meatloaf of a plane, with an aluminum arrow shaft for a tail boom. "Well, it's the 21st Century, I bet that carbon fiber is even better than aluminum." So I scream, "Turn right!" at our driver, who in a panic-- probably thinking we were being chased by an angry pack of youths, who having seen two girls and only one boy in a truck, decided to chase us down out of jealous rage-- whipped a giant F-150 across three lanes of oncoming traffic to bounce to a halt in front of the hunting store.
Note to self: there are two magic phrases whenever you want help or stuff for free. 1) "Hi, my name is xxx, from yyy University."; followed quickly by 2) "I've got a weird question to ask you." 5 minutes later, I walked out with an armful of carbon fiber arrow shafts that had had the feathers or tips broken. (Later on, when we were running low on LN2, starting a telephone call to the UT San Antonio chemistry dept. with "Hello, this is Dr. Kenneth Sebesta" was another magic phrase that was useful. I knew that title was going to come in handy some day!)
Wally-world was a bust, so getting back to the cabin, I set to work with the tools I had on hand. In the 35C heat. Using some clamps I found in the Batmobile (the large F-150 + trailer), I fashioned a jig that would hold my rods exactly 40cm apart. After a few abortive tries, I came across a winding strategy that would work. A couple hours later, I had the start of a frame. It basically looked like a tic-tac-toe grid made out of carbon fiber.
The motors were attached to some FRP that I had brought down for just this sort of tinkering, and then the FRP was attached to the booms via zip-ties, with some sticky double-sided tape added to keep them from walking down the beam. It's not what I would do again, given the choice, but as I had to work with the materials at hand it was all I could do.
2011-05-24_12-45-23_472.jpg 1.07MB 147 downloads
After having built the square frame, I noticed that the joints and small carbon fiber rod was nowhere near rigid enough. I strengthened it with two arrow shafts cutting diagonally across the center, which nicely provided a crossing place where I could mount the CC. However, note that one arrow shaft was attached to the top of the 10mm rod, and the other to the bottom. Thus, there was a 10mm gap between them at the center. This would prove a stumbling block.
The motors mounted and the frame completed, I had to figure out how to mount the circuitry. This was not easy, as as said earlier, the center mounting area was encumbered by the fact that the two arrow shafts where 1cm apart. However, right before I had left Boston, one of my lab assistants asked me if I would like to take along some funny little plates that you can zip-tie to, just in case. With a bewlidered look and a fatigued sigh, I'd said, "Yes." How wise he was and how lucky I was. Putting two of those plates together spanned the gap, with just enough space missing for me to slide in the fiberglass board where the CC was originally mounted on the batcopter v1.0.
2011-05-24_12-51-38_511.jpg 983.26KB 141 downloads
The battery was mounted on a bamboo boom behind the electronics. The fact that it off-balanced the QC didn't make a big difference, except that the rear motors had to work harder than the front ones.
2011-05-24_12-43-36_144.jpg 1.23MB 110 downloads
Finally, the netting needed to be added. This was done through the creative use of a hacksaw on the blue foam. That stuff is tough, but light. I need to figure out what it is and where it came from so I can buy more. The cut foam blocks were impaled onto the ends of the frame, and then kept from sliding around with some gorilla tape. Steel wire was passed from the front to the back of each major rail, and in the middle supported by some bamboo. The approach worked perfectly, both protecting the QC from shock and providing a large structure for the netting to rest on, so it wouldn't get pushed into the props by the bats.
2011-05-24_12-44-04_634.jpg 1.05MB 122 downloads
Finally ready to fly! Now just to tune it and set out to conquer the skies. Fast-forward over a particularly nasty calibration bug, which had already been fixed before I set out but hadn't yet made it into the main branch the day I got my last git checkout, I finally got the QC flying, and flying well. I estimate that I shaved over 500g off the final flying weight.
2011-05-22_16-13-56_593.jpg 1.03MB 130 downloads
Come that afternoon, we set out to the caves. Every day, we headed out at 3PM, into the hottest part of the day, to haul out several hundred kg of computers and cameras. At least this day I had an excuse not to help, I was too busy getting the QC ready for flight in the wilderness.
I had planned to fly without the GoPro and GPS, in order to test it first, but I was feeling frisky and so in the five minutes before the emergence started, I suddenly threw myself into a rage of action, pulling everything apart in haste in order to stick on the GPS and camera. (That haste is probably part of the reason why the netting failed on the second flight, but it was worth it.) As the trip leader yelled, "Kenn, it's starting!" I was wrapping the netting back on the QC. 1 second of bottom right stick and the QC was armed and ready to go. Up into the air I went, a little uncertain of myself, of my machine, and of the weather conditions at the top of a gusty bluff, with wind coming from behind the hill crest, catching everything in occasional rotors and down drafts.
After a few tentative motions, I realized that Quady the Batcopter flew like a dream. Into and out of the emerging swarms I went, each time going faster and faster. I eventually got to the point where I was swooping in from behind, flying faster than the swarm, catching up to the bats.
At one moment, a hawk came into the formation, and I chased it, too. (Unfortunately, out of the field of the cameras, so we couldn't see how it responded.)
I won't draw out the drama (I'm not particularly good at it anyway), so suffice to say that three major things happened:
1) A rotor got caught in the netting and I crashed inverted into the hard-scrable Texas wilderness, buckling one of the smaller CF rods. Well, nothing to do but keep flying.
P5230025.JPG 750.72KB 161 downloads
2) I crashed a second time in a spiraling descent. I called it quits after that one, as I hadn't understood why the second crash had occurred and feared an overheating motor or ESC as the culprit.
3) Somewhere, I got this magnificent infrared photo. Science meets science.
Batcopter_background.jpg 97.15KB 129 downloads
There's lots of video, but considering that each minute of video takes one hour just to move off the RAID hard drive onto an external drive, you can understand that this is not something that can just be uploaded to vimeo. I'll get some here just as soon as possible, especially of the crashes.
The next day, I decided that I wanted to treat you guys to a second example of a totally busted up quadcopter being flown fine by CopterControl. The video is here. The only reason it crashed was because the rotor caught on one of the delaminated carbon fiber strands. This time, the crash permanently disabled Quady the Batcopter, breaking its spine where the two arrowshafts met. Oh, well, it's another trophy for the "Version Wall" at the BU UAV lab.
The conclusion? Well, for the research, you'll have to wait a few years. But for the OpenPilot, it was a rousing success. Sure, there were a couple peccadilloes, a couple moments of frustration, a couple things that will need to be improved sooner or later, but the proof is in the above. In the hot Texas sun, with nothing more than the parts and tools that would pass the TSA gropers, and in the time of an afternoon, a complete quadcopter was born from the ashes of an older one, and flew to great success. Without the excellent CopterControl hardware coupled with the sublime GCS, this research trip would have been a failure, yours truly would have left Texas empty-handed, the grand experiment a bust. A most sincere, heartfelt thanks to all the developers who made this possible, but most especially Dave, who has done his best and sacrificed so much to make his dream a reality. Words do not express.
Your faithful correspondent,
Dr. Kenneth Sebesta
Université du Luxembourg
P.S. Flying commercial with a quadcopter strapped to your back is a great way to get all the girls to talk to you.
After all these years, I finally found a way to get girls to talk to me.
Edited by Kenn Sebesta, 09 June 2011 - 04:55 AM.