Replies to This Discussion

3D RoboticsPermalink Reply by Chris Anderson on August 28, 2011 at 5:27pm

The nav tuning guide is in the manual here.

• ▶ Reply

Permalink Reply by Justin on August 28, 2011 at 5:32pm

Sorry I should have mentioned that, I've been through the guide,
It did not help reduce the snaking.

I started with mucking around with values for auto stabilize
then moved onto the nav_roll stuff.

Circling: OK.
Maintaining altitude in a constant turn: OK
Stabilize: reasonably OK

however all the GA planes I can't get them to maintain a bearing without constant rolling back and forth.

But the RC plane worked with just whatever values I had left in there, so it seemed much less sensitive.

Do you have a known good set of values for any single X-Plane GA aircraft handy? Cessna, whatever .. then I can just load them or enter them and see..

• ▶ Reply

3D RoboticsPermalink Reply by Chris Anderson on August 28, 2011 at 5:41pm

You changed Nav_Roll I, Nav_Roll P and Crosstrack gain and it did nothing? Really?

• ▶ Reply

Permalink Reply by Justin on August 28, 2011 at 5:55pm

It did lots of things, for sure, it changed the snakiness, and the aggressiveness of corrections.

But I could not locate values that showed much overall improvement, and worse if I looked at the "servo outputs" especially if you started to add some "D", there started to be big instantaneous spikes in those as well.

Also, APM would only show updates to nav_roll once per second so I was never sure about what was going on at 50hz on the board.

 

BTW you can see this just with the default values, and any X-plane general aviation demo aircraft. I think.

 

What exactly is Xtrack Gain (degrees per cm)? 

Is there an explanation of why it would be necessary to adjust this and what direction and how much to adjust it by?

• ▶ Reply

3D RoboticsPermalink Reply by Chris Anderson on August 28, 2011 at 6:01pm

Telemetry rates can all be set in the Mission Planner:

 

• ▶ Reply

Permalink Reply by Justin on August 28, 2011 at 6:15pm

I figured that XTRACK GAIN has to do with how aggressively to follow the line between waypoints.

When I turned it off for GA, set to zero, all the snaking vanished and it made bee line from wherever it was to the WP.

When I turned it on but much lower than the default, I got the plane to follow the yellow line ok, but not snake.

I think XTRACK_GAIN is something that needs to be small for fast planes and large like the default for slow planes.
That was the missing piece of the puzzle. None of the PID settings needed much if any tuning really.

thanks.

BTW, there is some current issue with telemetry and the screen updating in APM: In the "Flight Data" Tab, in "Status", irregardless of the Telemetry rate settings, the numbers are not updating very quickly. Also in the green/blue fake HUD, GPS switches back and forth betweeen "GPS:No Fix" and "GPS: 3d Fix".

Meanwhile the dos console is spewing out:
"string to screen NoError" and "string done" text messages as fast as it can..

thanks.

• ▶ Reply

3D RoboticsPermalink Reply by Chris Anderson on August 28, 2011 at 10:50pm

Justin, I checked with Doug Weibel tonight (he's the main nav author) and he says by far the most important PID to work on first is Servo_Roll_P, which basically determines how much the control surfaces move.

The instructions in the Tuning guide are pretty good, and will walk you through this. I'll try to flesh it out in a bit more detail this week, too.

http://code.google.com/p/ardupilot-mega/wiki/Tuning

• ▶ Reply

Permalink Reply by Justin on August 29, 2011 at 12:23am

fair enough.

but changing the XTRACK setting was dramatic. I did a lot of experiments sitting here watching the x-plane fly around Innsbruck :)

 

I am pretty sure that setting has critical dependance on the flying speed of the model: small RC planes: 100 is ok. Big fast planes (or RC jets etc), must be much smaller. Perhaps you can ask him that? It is easy to see, you just get the plane (in x-plane) travelling between two waypoints and change dynamically the xtrack value, watch the snaking vanish with low numbers. And get really bad with 100 or above numbers (if the plane is ~80 knot + type plane).

 

I realize fixing the servo PID is necessary first, and nav roll pid next, BUT decent control is pretty forgiving of mistakes there: it'll fly the plane in windless situations without any drama with all kinds of values.

• ▶ Reply