PID parameter not exactly the same when Loading from a file or the copter

After much testing (especially in high wind simulations) I realized that one of the biggest issues with proper path following is not the PIDs themselves, but the fact that Ardupilot waits until it reaches the waypoint before it makes a turn, resulting in the flight path "bulging" out past the waypoint and the plane having to work its way back to the intended track.

This "bulge" is where many of the problems start to arise, snaking, overshoot, and in high wind scenarios it often results in missing the next waypoint completely as the plane can be blown several hundred meters off track before it has a chance to even finish its turn.

Here is a blog post that seems to have a great solution to the problem, with examples and source code:

http://diydrones.com/profiles/blogs/orbiting-and-turn-prediction

Although the above solution seems relatively complex, although arguably an ultimate long-term goal, wouldn't it be quite simple to implement some sort of basic "turning radius" setting that Ardupilot than took into consideration when making its turns? If the turning radius is set to 60m in its most basic implementation it could just start turning 60m before the waypoint. Slightly more advanced would be to calculate the exit angle and start turning 60m before a180degree turn, 30m before a 90degree turn, 15m before a 45degree turn, etc...

I believe having the plane leave the waypoint turn at least near the intended track would go a long way to making PID tuning much easier and "accurate" path following an actual reality, while both improving efficiency and flight time.

Replies

Mike October 25, 2011 at 8:49am

After much testing (especially in high wind simulations) I realized that one of the biggest issues with proper path following is not the PIDs themselves, but the fact that Ardupilot waits until it reaches the waypoint before it makes a turn, resulting in the flight path "bulging" out past the waypoint and the plane having to work its way back to the intended track.

This "bulge" is where many of the problems start to arise, snaking, overshoot, and in high wind scenarios it often results in missing the next waypoint completely as the plane can be blown several hundred meters off track before it has a chance to even finish its turn.

Here is a blog post that seems to have a great solution to the problem, with examples and source code:

http://diydrones.com/profiles/blogs/orbiting-and-turn-prediction

Although the above solution seems relatively complex, although arguably an ultimate long-term goal, wouldn't it be quite simple to implement some sort of basic "turning radius" setting that Ardupilot than took into consideration when making its turns? If the turning radius is set to 60m in its most basic implementation it could just start turning 60m before the waypoint. Slightly more advanced would be to calculate the exit angle and start turning 60m before a180degree turn, 30m before a 90degree turn, 15m before a 45degree turn, etc...

I believe having the plane leave the waypoint turn at least near the intended track would go a long way to making PID tuning much easier and "accurate" path following an actual reality, while both improving efficiency and flight time.
Developer

Michael Oborne October 24, 2011 at 6:06pm

there was ithink one version where the param files got messed around. make sure you use the latest, as this was fixed quite a while ago.
druckgott October 22, 2011 at 9:00am

I also have the second Problem on .82, I now have .84, but don´t test it again to load the parameter.

The problem was what if it saved 0.4 and load again it was 40.00 or 400.00 I can´t remind me, but not all parameter.

If I connect nextone, I will test it on .84

And I also have the first problem. But I think it´s not so importen to fix this

This reply was deleted.

Discussions

PID parameter not exactly the same when Loading from a file or the copter

Replies

Activity