Orbit 2.0.0.129
I have tested this both in Video mode and in timed interval Photo mode. I first observed that the orbit seemed notchy while doing a hyperlapse the other day. So I tested again today in Video mode and made a comparison orbit using Go 4.
Go 4 shows a very smooth constant yaw, while the DL OTF orbit seems to be almost flying short straight segments and then yaw back to target This is much the same way Go4 acted over a year ago before DJI fixed it with updates for the M2P. I did not test Orbit on the "127" build so I don't know if this is new behavior.
Video comparison: https://vimeo.com/488727181
Other observations:
With this version "129" I could not get Path, Circle or Spiral to continue past the setup screens. Tried multiple times on two different days.
With "129" Airdata does not show coordinates or any other information on the flights except the date and time. "127" shows up as normal.
Comments
3 comments
Video link is broken.
Link is OK.
129 had a bug with the functions not working, which is corrected in 130.
As for the performance, the Orbit component does break down the flight path into thousands of tiny samples (each of which may only be a few inches long), but I don't think that is actually the issue. There is error correction code in the flight controller that detects when the nose of the drone is different from what we want, and applies additional rotational velocity to get it back on track, which seems to be worse on drones with less capable gimbals, smaller airframes, or worse connection methods (think, mavic mini). We added Kalman filters a few builds back to try to "smooth out" how aggressive the error corrections are, but in some cases it seems like the error can become so large that it has to apply a decent velocity to get it back on track.
Having said all of that, we implemented Orbit mode in a different way that the Orbit component. Basically, we calculate the implied rotational velocity given the radius and horizontal speed, and just send a constant rotational velocity command. We did this out of a hope that it would fix the issue seen in your video, but in the end, we still had to add error correction into the algorithm because the nose of drone would just get off track (due to wind gusts or whatever else) and we ended up back at the same place. The only actually solution may be to go onboard (super low latency), which is what DJI is doing.
Please sign in to leave a comment.