Building a Custom Drone for University Agricultural Research: Feasibility, Ideas, and Feedback

Hello Ardupilot community,

I am a graduate student in a cranberry genetics laboratory who is putting together a proposal to purchase a lab drone for research (and attach a multispectral camera sensor). I have put together a list of options for my PI to review, with a range of customizable options. The least customizable options include a DJI Inspire 2 drone with hardware designed to mount to its gimbal interface and a DJI Matrice 2 drone with a newer Skyport interface that seems to be supported by many of the agricultural multispectral sensors I’ve been researching. A middle-of-the-road customizable solution is a DJI Matrice 100, which appears to be a good developmental platform, although adding components (such as live video streaming, visual guidance system) can quickly push the price skyward, and I am limited by proprietary DJI hardware and the fact that as an older model, DJI’s support for this model may dry up in time. At the most-customizable level, I’m proposing a hexacopter frame kit with suggested propulsion system, a gimbal that can attach ‘lightweight’ DSLR cameras and/or other sensors, with what I think is a nice-looking, high-quality flight controller unit that supports Ardupilot software (EMLID Edge).

I am pitching/recommending the custom-built drone option, foremost as an upgradeable and flexible platform, a non-proprietary hardware and software solution (less concern about repair or annoying software licenses), and significantly reduced cost for parts (new and replacement). However, my PI has said that he doesn’t want me to spend a tremendous amount of time assembling/engineering/debugging the drone before we can even get it into the field to collect data. Although I feel confident in my software background and skills, and I see that this community has built a solid framework around supporting various autonomous/remote-controlled vehicles, I am open to admit that my mechanical and electrical skills are limited. I’m hoping that with an intelligent selection of custom, high-quality parts/components, the assembly and off-the-ground debugging time will be minimal.

Thus, I wanted to know what the community thinks is reasonable in terms of doing this? Is this unreasonable? If not, what are recommended, high-quality controllers, frames, camera gimbals, etc. that can be reliably assembled with minimal debugging effort? What are good online sources for parts? Recommended frame kits? What are tradeoffs for hexacopter versus quadcopter design (I like that hexacopters can potentially carry more payload, and have some failsafe redundancy over a quadcopter in event of propeller failure, but the battery capacity requirements appear to be large for any reasonable flight time, e.g., 20-30 minutes)? I have read that one shouldn’t venture in building something more complicated than a quadcopter if experience is minimal in custom-built drones, but is this constraint reasonable with a purchase of good parts? In other words, with diligent selection of high-quality hardware, can hardware/mechanical debugging be kept to a minimum (again, I’m not concerned about software – I am very comfortable in this arena).

Current custom-proposal (feel free to give critical feedback on what I’m missing, what isn’t good, what’s overdesigned, etc.):

Thanks everyone! I look forward to engaging with this community, and at a minimum, I hope to build my own personal drone.

Agreed you aren’t going to be able to do what you’re describing with M210 - if that’s what you mean by matrice 2. I have a m210rtk and its’ true that DJI has customizable SDK but that requires you to be in their enterprise program (may be called something else) which is very expensive and still hugely restrictive in terms both of capabilities and hardware options. You’ll get stuck with very neat and expensive DJI payloads that won’t work with newer (DJI) platforms and “smart” batteries which have the universes’ buggiest and most unnecessarily annoying firmware and which cost 3-5X market. Fancy DJI monitors (crystalsky) or controllers (cendence) won’t work with other DJI platforms, like the Mavic, while non-DJI sub $100 android tablets work with Matrice & Mavic both.

Use Ardupilot or PX4 and pixhawk for greatest support, documentation, customization/configurability, reliability… Add companion computer for higher level and state-of-art applications

Thanks for the feedback, and I’m glad my concerns about hardware/software proprietary issues surrounding DJI. We do have a DJI Phantom 4 Pro drone, which is fine, but I’ve been advocating for expanding the sensor capabilities and this led me down the path of advocating for an entirely customizable, better product.

Just out of curiosity, do you not like the Edge flight controller because it is very expensive, or are there other aspects that get in the way of being flexible? It looks like a high-quality product, but without any background on my end, it could just look slick and be a piece of junk.

Yes, I’m in US. If I go with Pixhawk, what do you recommend for additional hardware that is provided by default in the Edge kit (2 IMUs, barometer, GNSS receiver, HD video transmitter, etc.)?

Oh, and one last thing: You mentioned the Rededge sensor in your work. I’ve seen the Micasense Rededge MX model, which is in my list of potential commercial multispectral cameras to consider, but I am concerned that some of these sensors are also tethered to proprietary software/hardware interfaces. You’ve found the Micasense product lines to be fairly open and standard in software and hardware interfaces?

Thanks for the information on the costs associated with the DJI SDK. I never even considered that.

I am building Drone with similar requirements. find the parts order to build the same.

Hey, anyone with very specific requirements is having a hard time picking the right drone. Buying quality parts is a given, otherwise you multiply your troubleshooting time by a million. If you go the custom drone route, even if you have a guy who’s built them before, you’ll likely spend a lot of time debugging the copter and all of the sensors. Personally, getting these things to perform reliably was my biggest headache. I started from scratch and 6 months later I had build 2 drones, one of them being Pixhawk/Ardupilot with a Sony A6000 camera and no gimbal.

Your other option is to avoid all of this headache and pick a problem that can be solved using a DJI product or another proven drone out there (not all that many I feel like). Best of luck!

I also think that an UAV for an specific task, such as agriculture or research or data analysis should be custom made. It is the best way to ensure a perfect lay-out of the sensors and payload, and the perfect match of other requirements (time of flight, AUW, gps, indoor capabilities, etc.)

In my case I bulid the frame, sensors supports and motor holders with a 3D printer. It is quite resistant, as it has supported some bad landings, but in case of a hard accident I can re-print and install the parts with in one day.

With some experience, building and sensors configuration times are short. I made a complete rebuild of my drone in two days (an Hexa with a lot of sensors), but I know people that has built a small quadracer from scratch in about five hours.