alt Hacker NewsI was unable to ever get it to fly reliably without GPS. It was probably stupid to drop $7K on drone GPUs and all kind of gadgetry (6 battery bay for rapid charging, etc), but it was just really really hard to pilot around in Maryland (Montgomery County). I would constantly have it throw up warnings and alerts, even only hovering a few feet above the ground for small scale testing. I would have to disable the GPS to do small scale testing, and then with GPS enabled, it would straight up not allow me to pilot it. When I moved to Miami, brought it down there, but I managed to find an apartment right smack-dab in the MIA no-fly zone as well. The smaller drone was allowed to fly though, so I eventually got a small Mini 2 IIRC, which was a lot easier to pilot, but I was just so disappointed in not being able to use the larger scaled up version. I wanted to do realtime facial recognition (not at scale, just to show that commercial drones can be turned into research demonstrators) on the onboard GPU (apparently just a NVDA Jetson from 2017 era)
The irony is the M100 is genuinely great hardware - the payload capacity, the SDK access, the flight time with extra batteries. But DJI's geofencing treats the entire DC metro like a no-go zone, which makes sense from their liability perspective but means the thing is basically a $7K shelf ornament unless you want to deal with LAANC authorizations for every single flight.
Gah that sucks. I've looked at the hardware specs and basically ended up drooling over it and realizing that my homebrew stuff will never be able to compete. But the optics alone on that DJI stuff is nothing short of science fiction compared to what you can put together on a hobbyists budget. But for $7K you can build an octocopter with twice the range and twice the payload, which may not be as impressive on paper but can be pretty useful as well.
The larger agricultural drones are also amazingly impressive, those I've seen up close doing real work and they are so reliable it is almost boring.
I wonder what the reason is that yours behaves the way it does, that sounds like a real challenge to find out though with the closed system like that.
Drones that rely on GPS are very iffy as soon as the GPS fails, I've seen more than one inexplicable 'fly-away' happen. I've found a really neat trick to test drones that are not 'known good': just find yourself a long stretch of really light chain and tie it to the drone. As long as it behaves: no problem. But if it tries to take off by itself at some point the length of chain weighs more than the drone can handle and it will stop ascending. That way at least you have some kind of safety measure that does not immediately impact the drone in a material way as long as it is near to you.