Re: [Paparazzi-devel] ground / airborne interaction

[Oswald Berthold]

hi felix, thanx for commenting. some more text inline below ..

cheers, opt

Felix Ruess writes:

> 1. That is correct, the integrated error is currently not reset when gains
> are updated via settings.
> However that could easily be added in settings with a "handler" that will
> reset it when a setting is changed.
> Not sure if we really want that to happen every time a gain is changed
> tough...

it could be fully reset or clamped to a reset value, along the lines of
what's listed here

https://en.wikipedia.org/wiki/Integral_windup

> 2. Hm... good question... Sending RC_4CH messages (what joysticks do) and
> would be one solution, but I think you currently can't use a regular RC and
> joystick at the same time.

so i could just use the joystick, and mix manual and programmed commands
in my script and send the sum via RC_4CH.

> I think the long-term clean solution would be to set the quad into a sort
> of external guided attitude mode and send attitude commands via separate
> datalink messages.

hm, how do you handle the mixing of attitude control commands from RC
and the nav/guidance system, isn't that the same/similar situation?

> 3. To try to find out where the message is dropped, could you try with
> connecting a usb<->serial cable to the telemetry uart directly and see if
> you also don't get the setting confirmation?

ah, should've thought of that, i'll try it.

> On Wed, Sep 23, 2015 at 4:52 PM, Oswald Berthold <address@hidden> wrote:
>
>>
>> hi all,
>>
>> i spent the last days implementing this PID tuning approach [1] for a
>> pprz system, see [2] in case you're interested. basically it all works
>> but there are some items left which i'd like ask you about:
>>
>>
>> 1. i have the impression from observing the behaving copter that the
>> error integral doesn't get reset when i update the params (pgain_,
>> dgain_, igain_, ddgain_). is this the case and how would i best go about
>> it: send a reset request msg or just put that into airborne code?
>>
>> 2. currently i'm using the RC to manually send the probing motion
>> commands during evaluation but i would like to send a fixed control
>> signal from my python script. should a just pretend being a joystick?
>> do joystick and RC work together?
>>
>> 3. i'm kind of confused about the settings confirmation: basically all
>> settings get sent at about 1 Hz in cycle mode. when you do a dl_setting
>> the AP responds immediately with a confirming dl_value. the confirmation
>> msg though seems to get dropped on my telemetry connection quite often
>> though. i gave up on trying to wait for the confirmation and just
>> transmit the params at 1 Hz. i already replaced the 57kb xbee S1
>> modems with 115kb rctimer ones, which improves things a bit but i still
>> get dropped msgs. any ideas on how this could be done properly?
>>
>> (i know i could figure out 1 + 2 by surfing the code but if anyone can
>> provide hints that'd help).
>>
>> cheers, opt
>>
>>
>> [1]
>> http://www.cs.unm.edu/~afaust/MLPC15_proceedings/MLPC15_paper_Berthold.pdf
>> [2]
>> https://github.com/x75/paparazzi/blob/x75/sw/ground_segment/python/paramopt.py