mTECS

[LorenzMeier]

This is Thomas' total energy control system. Heavily tested in the last days, its now a good time to start ironing out the last kinks and establishing trust logs. Should also be made the default on master.

@thomasgubler @sjwilks @AndreasAntener Please provide additional comments here. Less branches will mean less confusion in the next weeks testing 8).

[achambers16]

I'm interested in helping to test this branch. A few questions:

• Should I wait for Navigator Rewrite #1046 to be merged in?
• Is there any documentation on mTECS? How about mTECS versus the current TECS? Besides finding bugs, what should I be looking for in terms of performance or behavior?

[thomasgubler]

Should I wait for #1046 to be merged in?

This here is a merge of ongoing mtecs, navigator and estimator work, highly WIP https://github.com/PX4/Firmware/tree/navigator_rewrite_estimator
You can test mtecs separately though

is there any documentation on mTECS?

Currently only a diagram and the param descriptions (in the code), more to come of course
http://pixhawk.org/dev/control/mtecs

[LorenzMeier]

@thomasgubler When can we expect some more mTECS docs? I think you have now a whole bunch of developers asking for them and it really would benefit the further development to make a grey box out of the current black box. We're not quite far from merging it to master and so documentation should be a primary concern now.

[thomasgubler]

As soon as most technical issues from the weekend testing are resolved...

[thomasgubler]

Meanwhile I added more docs http://pixhawk.org/dev/control/mtecs
Also airspeed is now filtered (additionally) for use in the mtecs calculations.

[achambers16]

Two recommendations here:

1. Publish the whole 2-by-2 covariance matrix for windspeed

• Might as well capture and use all the data that we have. Easier to put it in now.

2. Set to negative number for unknown or not estimated

• While unlikely, zero variance is a possible value for variance calculation. Negative variance is not possible. Let's set a good convention.

[thomasgubler]

This goes into the same directions as my comments here #1036
--> Let's extend the interface when we need it

[LorenzMeier]

@thomasgubler Compared to the state of master this is by now in a much better shape - given that better docs than for master exist, how about getting this in?

[achambers16]

I'm not having very good performance in HIL using X-Plane 10 using default
parameters. Maybe I need to tune it? I'm using the "HilStar, X-Plane"
airframe configuration.
I'm seeing quite a bit of oscillation in airspeed and altitude. And poor
airspeed control to the FW_AIRSPD_TRIM setting.
Disabling mtecs via the parameter results in better flight performance.
Any ideas?

On Thu, Jun 26, 2014 at 8:50 AM, Lorenz Meier notifications@github.com
wrote:

@thomasgubler https://github.com/thomasgubler Compared to the state of
master this is by now in a much better shape - given that better docs than
for master exist, how about getting this in?

—
Reply to this email directly or view it on GitHub
#1037 (comment).

[LorenzMeier]

@achambers16 The params for the HilStar are not tuned at all. You will want to tune them using the guide Thomas wrote:
http://pixhawk.org/dev/control/mtecs

[thomasgubler]

The only thing preventing this from merging right now is the fmuv1 flash size issue.

[LorenzMeier]

@thomasgubler Please see #1084 and #1085 for flash improvements. I propose to merge both and to merge this to master.

[thomasgubler]

@LorenzMeier do you want to keep the tests removed for fmu1? They still don't fit...

[LorenzMeier]

@thomasgubler Given that your comment is addressed, merging this now. This is a really great contribution and a lot of work, thanks!