paparazzi-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Paparazzi-devel] Quad Tuning?


From: Gautier Hattenberger
Subject: Re: [Paparazzi-devel] Quad Tuning?
Date: Sun, 18 Jan 2015 22:07:16 +0100
User-agent: Mozilla/5.0 (X11; Linux i686; rv:31.0) Gecko/20100101 Thunderbird/31.3.0

Hi,

Actually, I can reproduce it (and even worse in fact). The source of the problem is that a function tries to update the waypoint name in the flight plan when it is changed in the list. In fact only when the 'wp' or 'from' attribute is used.
One of the problem, is that the callback is triggered for any changes on the GTK tree and this happening a lot without telling much if it is just some attribute modification or or moving a node. When you start changing the WP order, it becomes a big mess.
The solution to fix this is absolutely not obvious. Although it is a very nice option, I'm not really sure anyone ever noticed it, and it is not changing the occurrences of 'WP_XXX', so it doesn't solve all problems. We even know now that it is creating a lot of problems.
So I would recommend, to remove this option.

Gautier

Le 18/01/2015 02:34, Felix Ruess a écrit :
That is so weird, I did exactly the same (using the same version) and it doesn't happen...
Could you please do a "make clean && make" and check if you can still reproduce it (since then it really would be the same state I tested)?
Maybe @gautierhattenberger has some idea about what could cause this?

On Sat, Jan 17, 2015 at 9:19 PM, Ori Pessach <address@hidden> wrote:

On Sat, Jan 17, 2015 at 6:36 PM, Felix Ruess <address@hidden> wrote:
Yeah, that would definitely help.

On Sat, Jan 17, 2015 at 5:25 PM, Ori Pessach <address@hidden> wrote:

I haven't made any changes to the GCA code. The local changes are too the vision code, and various XML files.

I'll try to make a video to show you exactly what I'm seeing, if that will help.

On Jan 17, 2015 8:53 AM, "Felix Ruess" <address@hidden> wrote:
It means that the version you are using is 203 commits after the v5.3_devel tag, the git sha1 begins with aafaa6d and the dirty flag says that you have modified files/uncommitted changes in your tree.

And of course I'm saving the file and looking at that, otherwise you wouldn't see any changes at all.. ;-)
Tried with v5.3_devel-203-gaafaa6d, can't reproduce it there either...
Do you have any local changes that could be relevant?

On Sat, Jan 17, 2015 at 4:25 PM, Ori Pessach <address@hidden> wrote:
Are you saving the xml file, or just looking at the UI? The problems don't show up in the UI, but the xml file is saved incorrectly.

Turns out that I'm running 5.3_devel-203-gaafaa6d-dirty. Whatever that is.

On Sat, Jan 17, 2015 at 5:14 PM, Felix Ruess <address@hidden> wrote:
Hi Ori,

Run ./paparazzi_version to get an exact version report. Same that is reported when compiling.

Hm.. I just tried with v5.2.1_stable-17-g3fbf056 and no matter what I do, it never changes the waypoint name in a block.
Maybe I misunderstood how you reproduce this, but dragging waypoints in the tree view just changes the order (as it is supposed to).

Felix

On Sat, Jan 17, 2015 at 3:44 PM, Ori Pessach <address@hidden> wrote:
I'm using Paparazzi 5.2.0 stable, at least according to CHANGELOG.md.

On my end, the problems are very easy to reproduce. The easy step to forget is saving the flightplan.

Ori

On Sat, Jan 17, 2015 at 12:48 PM, Felix Ruess <address@hidden> wrote:
Hi Ori,

I can't reproduce any of the problems you described....
The attributes are generally alphabetically sorted after saving with the flight plan editor, but that makes no difference.
Only weird thing I could manage by dragging was to put a waypoint in another one as a child, but that of course doesn't even compile.

Could you please check again and report the paparazzi_version you are using?

Cheers, Felix

On Sat, Jan 17, 2015 at 3:24 AM, Ori Pessach <address@hidden> wrote:
OK, here are the steps to reproduce the bug:

0. Make a backup copy of rotorcraft_basic.xml
1. Load rotorcraft_basic.xml in GCS, in edit mode. 
2. Open the Waypoints branch in the tree widget.
3. Drag the P1 waypoint Above the CLIMB waypoint.
4. Save the flightplan.
5. diff the new rotorcraft_basic.xml file with the backup copy you made in step 0.
6. Be happy you didn't try to fly with the edited copy. :)

Depending on how I drag things around, I can get the flightplan to break in different ways, including the problem I reported earlier (where CLIMB was replaced with P1.) I've seen other things getting clobbered, though. 

What's disconcerting about this is that the UI doesn't give the user any indication that something is wrong - the tree widget doesn't show the internal changes that get saved incorrectly. That's pretty mean. 

Ori


On Fri, Jan 16, 2015 at 6:20 PM, Felix Ruess <address@hidden> wrote:
Not unless you use the GCS as a visual flight plan editor...

On Fri, Jan 16, 2015 at 4:41 PM, Ori Pessach <address@hidden> wrote:
You're right. I've no idea how my copy ended up in this state. 'git diff' shows a lot of differences between my copy as it exists now and the one I checked out - mostly xml attributes getting reordered. Does GCS make changes to the flightplan file?

On Fri, Jan 16, 2015 at 5:19 PM, Felix Ruess <address@hidden> wrote:
Um, the rotorcraft_basic.xml flight plan in the repo is correct in that regard: https://github.com/paparazzi/paparazzi/blob/v5.4/conf/flight_plans/rotorcraft_basic.xml

On Fri, Jan 16, 2015 at 4:15 PM, Ben Laurie <address@hidden> wrote:
On 16 January 2015 at 00:40, Ori Pessach <address@hidden> wrote:
> Ben,
>
> I think I just figured out why my drone was doing what you described here
> while taking off in NAV: I was using the rotorcraft_basic.xml flightplan as
> a basis for my flightplan, and here's what the  Takeoff block looked like:
>
> <block name="Takeoff" strip_button="Takeoff" strip_icon="takeoff.png">
>   <exception cond="stateGetPositionEnu_f()->z > 2.0" deroute="Standby" />
>   <call fun="NavSetWaypointHere(WP_CLIMB)" />
>   <stay climb="1.0" vmode="climb" wp="p1" />
> </block>
>
> As far as I understand, this moves the CLIMB waypoint to the drone's current
> location, and then attempts to climb to waypoint P1 at a rate of 1m/s until
> the drone reaches an altitude of 2 meters, at which point control switches
> to the Standby block.
>
> The problem for me was that P1 was not near the drone's starting point, and
> the drone was trying to make a sharp turning maneuver while still on the
> ground, causing all sorts of ground encounter issues.
>
> Changing the stay element to:
>
> <stay climb="1.0" vmode="climb" wp="CLIMB" />
>
> fixed the issue I was having with takeoff, and allowed the drone to finally
> takeoff cleanly, climb to its target altitude and move to the Standby block.
> I also changed the standby block so that it stays at the CLIMB waypoint,
> instead of P1.

Haha ... that is awesome! Wish I'd noticed it, would've saved me a lot
of head-scratching. You should submit a pull request. Or I will if you
don't.

> This was the first time I was able to fly the drone completely autonomously
> with Paparazzi. Once the drone was off the ground, it seemed to navigate
> between waypoints and land perfectly fine. I have to say that flying is a
> lot more satisfying than crashing...

Sure is!

> On Mon, Dec 29, 2014 at 8:28 PM, Ben Laurie <address@hidden> wrote:
>>
>> Is there a good tutorial for choosing all the different PID parameters
>> for a quad?
>>
>> Mine flies fine when flown manually, but is perhaps a little soft.
>>
>> It also has problems taking off in NAV: it tends to ramp up the motor
>> slowly and when its got enough lift to take off it instead flips over
>> (I presume because its trying to head off in some direction other than
>> straight up?).
>>
>> Takeoff in ATT is just fine.
>>
>> Could this be down to PID tuning?
>>
>> (I also wonder if I've managed to break the GPS antenna - error is around
>> 6m).
>>
>> _______________________________________________
>> Paparazzi-devel mailing list
>> address@hidden
>> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>
>
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>

_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel


_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel


_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel



_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel




_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel


reply via email to

[Prev in Thread] Current Thread [Next in Thread]