APM Navigation Performance

I am trying to optimize the navigation performance of my APM skywalker for aerial mapping applications. I have just downloaded the latest version of APM as of August 14^th (2.211), hopefully this version includes Andrew’s bug fix (http://diydrones.com/profiles/blogs/fixing-a-magnetometer-bug-and-how-to-use-mavlink-to-diagnose-apm). I am using a magnetometer and have had periodic nav problems similar to Andrew’s, but the most recent flights looked good. I have done some basic PID tuning as per the manual, and get the following nav performance:This flight had about 3 knots of wind from the right side.

Looking at other recent postings, this performance looks fairly typical of APM. However, looking at some of the other autopilots such as the T3 R4 challenge, (http://diydrones.com/profiles/blogs/t3-round-4-map-a) the nav performance can be quite a bit better. Is this about as well as one can expect for the moment, or does anyone have suggestions on improving it? Has anyone made a video of how to optimize tuning and navigation of APM? Any tricks or secrets out there?

Second installment of APM navigation training using a skywalker. After the previous day's successful nav training flights, I had the opportunity to implement Chris's advice and work on the NAV-Roll PID's.
The first flight after having made no changes from the previous mornings flight yielded a magnetometer around 60 degrees out and would not converge (or some other issue) even after flying for 15 minutes. So I went into CLI mode and disabled the Compass. Then took off again to tune the nav without compass, and increased the NAV_ROL P value to 0.9 without much effect, then the NAV_ROL I value to 0.2 and this had a good effect. I played with the NAV_rol I value a bit but mostly stuck with the 0.2 value. The navigation quality was not much better than the day previously, but the wind was stronger and this flight had no mag compass.
After switching batteries, I decided to enable the compass again to see if it would converge, and voila, this time it did so after 3 to 5 minutes of flying, and now the nav was noticable improved. Thanks again for the nav-roll I suggestion, this has a quite significant effect on dialing in the nav quality.

test16aug2.png

test16aug2-1.png

test16aug3_wmag-sm.png

Replies

Robin Behrens August 23, 2011 at 12:41am

Hi, I also notices a similar patters and recently tried putting in extra waypoints. Although the way point may seem redundant becasue they are still in a straight line I find a straighter flight path.
Gustav Kuhn August 22, 2011 at 12:20am

This was my latest flight with a ST Models Discovery, very promising results :-)
Cliff Gill August 16, 2011 at 7:26pm

Second installment of APM navigation training using a skywalker. After the previous day's successful nav training flights, I had the opportunity to implement Chris's advice and work on the NAV-Roll PID's.
The first flight after having made no changes from the previous mornings flight yielded a magnetometer around 60 degrees out and would not converge (or some other issue) even after flying for 15 minutes. So I went into CLI mode and disabled the Compass. Then took off again to tune the nav without compass, and increased the NAV_ROL P value to 0.9 without much effect, then the NAV_ROL I value to 0.2 and this had a good effect. I played with the NAV_rol I value a bit but mostly stuck with the 0.2 value. The navigation quality was not much better than the day previously, but the wind was stronger and this flight had no mag compass.
After switching batteries, I decided to enable the compass again to see if it would converge, and voila, this time it did so after 3 to 5 minutes of flying, and now the nav was noticable improved. Thanks again for the nav-roll I suggestion, this has a quite significant effect on dialing in the nav quality.
test16aug2.png

test16aug2-1.png

test16aug3_wmag-sm.png
3D Robotics

Chris Anderson August 15, 2011 at 7:51pm

That's not bad, but you probably improve it with tuning. Maybe turn up the NAV_ROL_I and play with the Crosstrack gains:

This reply was deleted.

ArduPlane User Group

Replies