Drone design theory and drone construction methods?

[HansenJC]

Does anyone have suggestions for an active forum for builders who are working on medium to heavy drones 400mm to 600mm? (not minis, micros, or racers.) I do not see an area in this forum where others are working on building multirotor drones from scratch. Is this general forum the best place? I have joined DIYdrones.com and the discussions on that forum have faded away. I am looking for theory and practice from experienced builders.

 

[PatR]

Most heavy lift drones start at 600mm and go up quickly from there. For the most part everything under 600mm falls into at most a medium lift class.

As for a “builder forum” I don’t know there would be enough interest to carry it. Those that build, troubleshoot, create, and invent are a fast dying breed in multirotors, and even in RC model aviation. I’ll dare say those that can and do have vastly better command of their systems though.

 

[clolsonus]

Does anyone have suggestions for an active forum for builders who are working on medium to heavy drones 400mm to 600mm? (not minis, micros, or racers.) I do not see an area in this forum where others are working on building multirotor drones from scratch. Is this general forum the best place? I have joined DIYdrones.com and the discussions on that forum have faded away. I am looking for theory and practice from experienced builders.

If you do find any interesting places with people interested in building drones from scratch, please report back here. I've been watching out for this myself, but so far haven't found much. For many tasks, it's really hard to beat a phantom or mavic (price vs. performance). But if you do need to step out of the main stream to do something a little different, or need a specific improved capability for a project, there just doesn't seem to be large communities DIY'ing their own drones any more. You can still buy parts and frames and flight controllers and mobius gimbals, but where do you go to bounce crazy ideas off people and ask questions? Where do you go to find people with similar interests and capable of serious discussions? There must be forums or groups out there somewhere? Beuller?

I recently heard of a book that came highly recommended from a colleague called "DIY Drones for the Evil Genius." The first chapter summarizes a lot of basic aerospace theory and design rules of thumb, and then the book goes off into other specifics. I haven't had a chance to pick up a copy, but it might be worth checking out.

The rest of this is me just speculating. I don't know anything special. I know even less about the commercial side fo the world. I work at a university after bobbling along for a few years as a partly-successful entrepreneur (paid my mortgage every month, but eventually had to walk away and get a real job.) I guess rather be creating and building and flying anyway rather than sitting in a board room trying to answer angry investors and hold them off from pulling the plug.

So I don't know how much of this is cause and how much is effect, but when 3DR pivoted away from open-source hardware to a fully proprietary solution, they lost their motivation to lead and inspire the DIY community and I think that's when things started to take a nose dive. Somewhere in there the FAA introduced part 107 and made commercial drone operations legal. That also had a profound affect on the industry. No longer could companies exist on hype, they actually had to start turning a commercial profit doing real drone work ... and many companies ended up going away, or pivoting to mostly software development. Then DJI just kept killing it with pace of new products and functionality and quality and price so that (a) it made a lot less sense to build your own, and (b) it was really hard for other companies to compete. But now DJI is killing off the phantom (maybe? probably?) so if the mavic doesn't do what you need, then your options drop rapidly (and/or your cost increases rapidly.) But now you are in a tiny non-mainstream segment of the market and no one can hear you scream!

I sure hope the DIY community rebounds, but I'm sure it will never be what it was. I also feel pretty strongly that the fixed wing will scale far better than quads for many future uses. Quads will always be great for real estate, construction, and small-scale survey/mapping. But at some point, people in industries like agriculture, forestry, etc. will realize they need longer flight times and larger area coverage to make remote sensing work financially viable ... and that ultimately means larger fixed wing aircraft. Just my 2 cents.

 

[HansenJC]

I sure hope the DIY community rebounds... <snip> ...Just my 2 cents.

You have summarize my own observations in every regard. I may begin to acquire a few interesting books like the one you have mentioned. I will continue to post here in the general forum if there are no objections.

 

[clolsonus]

I know the DIY quad copter community has grown small. It seems like the DIY fixed wing community is just that much smaller even. Most of that sliver of remaining people interested in fixed wing uav's will just fly a pixhawk flight controller with px4 or ardupilot. But if there are any people left in the world that are interested in DIY'ing your own flight controller and learning more about that, we have an alternative system that is a lot leaner in many ways compared to the px4 and much of it is written in python. The goal is to be rock solid and as simple and accessible as we can be. PX4/Ardupilot projects are great, but they support everything for everyone and even though they are open-source, the complexity of their systems ends up being a substantial barrier for new people wishing to learn about these things and understand something about them. Our system is probably more complex than we realize too, but we have worked hard to avoid complex programming paradigms, kept the structure as flat and standard as possible, we avoid using a tricky threaded architecture, incorporate a teensy (arduino style processor) with a linux processor. But as I commented earlier, it's hard to even find people in the world interested in thinking about this stuff any more.

 

[PatR]

I would be interested in learning more about that FC. Pixhawk and PX4 are great but as you mentioned so broad in scope they make the programming process more difficult than it needs to be. Vector is much simpler but also more limited.

 

[clolsonus]

I would be interested in learning more about that FC. Pixhawk and PX4 are great but as you mentioned so broad in scope they make the programming process more difficult than it needs to be. Vector is much simpler but also more limited.

Hi Pat, my work with this system began before DIY drones and the APM1 going back to the mid-2000's. I've been at this a long time so I could pile on information and thoughts and motivations and stories for hours. That's hard to do well in a forum, so I'll say a few things and feel free to ask more specific questions.

I've always liked linux because of it's capabilities and tools and fairly low level connection to the system resources. Raspberry pi and beaglebones run linux. Way back when I started we were working with early gumstix boards which also ran linux.

I like simplicity. I think it is analogous to weight when designing an airplane. You can't always make things as light as you wish, but it's a battle that needs to be fought every day and any gains are immediately felt in increased performance or capabilities. I don't think it's too much different when coding.

I like python. When I port my C++ code to python, I get the same result at often 1/3rd the lines of code. Fewer lines of code is usually directly proportional to simplicity (assuming no one is going out of their way to write obfuscated code.)

I like figuring things out for myself. The reality is that the system is going to go fly a grid pattern or a circle for an hour+. If I slap someone else's controller and code in an off the shelf airplane, it's boring. The excitement comes from watching your own code do all the magic just the way you intended it. It doesn't always work out as intended and that makes the successes even more satisfying.

I like robustness. I have run my flight controller for more than 9 days straight flying a realistic flight simulation producing realistic sensor data (flightgear). I watch things like my flight controller memory footprint to make sure it never changes during this time. I make sure there aren't any glitches or bombs. If something does bomb or hang, it really stresses me out until I get to the bottom of it.

I like simplicity of design. I'll talk your ear off for another hour about threaded vs. not-threaded architectures. For a flight controller there is a clear information flow: read sensors, compute attitude estimation (ekf) and other sensor processing, compute navigation and control, send out actuator commands. I don't want these things happening out of order or on some non-deterministic time schedule. I've thought a lot about timing and end-to-end throughput. PX4 has a very elegant threading and messaging system (uORB) which is very beatiful from a computer science stand point. But from an engineering standpoint of producing time deterministic operations, it falls pretty short. I've heard they've made a lot of improvements here, so I'm not trying to be negative on them at all, just pointing out my own observations from a year or two ago when we dug deep into the PX4 architecture.

We use a big processor/little processor design. We have a little processor (teensy-3.6) that does all the sensor I/O and servo PWM generation. But this pretty much all it does. The big processor runs linux and our flight control app is a hybrid of C++ and python. The big processor does all the heavy lifting work ... computing the EKF, doing WGS-84 great circle navigation, mission, logging, communication, guidance, control. The result is two simpler systems working together (versus something like PX4 which is a big complex monolithic app.) Our little processor which handles sensor I/O also takes care of all our hard real time work. The big processor waits for sensor data from the little processor, crunches through the main loop, and sends servo commands back to the little processor.

We aren't trying to rule the world with this system; we simply have been developing it to support our various projects over the years. I have a lot of time and thought personally invested in it, so I am completely biased. Our lab also runs a few things with pixhawk and for student projects, pixhawk has worked pretty well. It's relatively easy to get something installed and flying and there is a lot of documentation. However, we are primarily a research lab, so having an in-house developed system that we understand from top to bottom and can modify for our needs ... that has really proved beneficial. PX4 is open-source, but the sheer complexity is a steep hill to overcome. We also have an extremely complex project with 12 independent flight control surfaces, 2 EDF motors, and retractable landing gear. In addition we have several IMU's and analog sensors spread across a 14' wing span. This just ends up being beyond the capability of a standard pixhawk. Our system is expandable so we actually have 3 additional teensy processor boards that run in lockstep with our little processor. The little processor aggregates all the sensor information and hands it to the big processor all at once. This is way more complex than our typical projects, but we had to design something ourselves to support the large number of sensors and servos and simultaneously meet strict deterministic timing requirements.

All that said, in it's typically configuration our flight controller is a beaglebone black with a simple cape that has a teensy3.6 onboard. The cape mostly just routes connections between everything ... it's not a complex board. One of our guys spun off a company to do flight controller hardware (and other stuff.) All our code and research is licensed with the MIT open-source license, but hardware is hard ... and it just costs money for someone to source the parts and build them up. You can google bolderflight to see what the hardware looks like.

One of the students in our lab integrated our flight controller into an x-uav talon for an entomology (insects) project and made a little video. If you have a couple minutes, it is a good little overview of the sorts of things we have been doing. (apologies if I've shared the link here before ... I'm getting old and starting to repeat my stories.)

That's probably way too big of an information dump for a forum post, but there is a lot of details and aspects and thought and experiences that have gone into all of our work. It's hard to know what specific questions people have or what angle they might be interested in. Everyone is interested in slightly different use cases, so I think it would be really empowering to have a community somewhere that is capable of building up flight controller hardware and software from scratch. DIY drones used to be the goto community for this, but now ...??? If I haven't just completely scared everyone off, please feel free to ask more specific questions. I don't know all the answers, but happy to share what [I think] I know.

 

[PatR]

What you reference captures my interest enough to suggest we go into private messaging. There’s a couple people I think you should talk to. One is a PHD at one of Boeing’s UAV off shoots. Another owns a private company that creates much of their own hard and software.

I’m not a coder, just a user, but I understand what you were telling us. You caught my interest hard and fast when you mentioned Raspberry pi and Gumsticks. Lotta power in small packages.

 

[clolsonus]

I’m not a coder, just a user, but I understand what you were telling us. You caught my interest hard and fast when you mentioned Raspberry pi and Gumsticks. Lotta power in small packages.

From my perspective, the really nice thing about linux and other open-source systems is they can weather the whims and storms and hype of the commercial marketplace and just keep chugging along at a steady pace. I've been running Linux as my main desktop environment since approximately 1993 or 1994. For me it is very nice to have all the same libraries and tools on my flight computer, it's great to have python support with it's entire eco-system. Like you say: lotta power in a small package!