paparazzi-devel
[Top][All Lists]
Advanced

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

Re: [Paparazzi-devel] BODY_TO_IMU Rotation for Hovering VTOL Fixed-wing


From: Bart Remes
Subject: Re: [Paparazzi-devel] BODY_TO_IMU Rotation for Hovering VTOL Fixed-wing
Date: Thu, 23 Feb 2012 00:48:08 +0100



On Thu, Feb 23, 2012 at 12:15 AM, Chris Wozny <address@hidden> wrote:
Bart,

Are you using quaternions internally for the control system now or
still Euler angles?
for initialisation we still use the Euler init (resulting in this fold behaviour    http://www.youtube.com/watch?v=5kzFLrNgC9c  we have to keep de Z axx axes neg when initializing resulting in the correct initialisation http://www.youtube.com/watch?v=0i2m6q3KyT0)
we will test the  quaternions initialisation of felix next week when we had the live demo of the uavforge competition.
but this will resolve the issue. so please test this and use this!!!

when flying the quaternions ate used in ahrs and stabilisation
    <subsystem name="ahrs"          type="int_cmpl_quat"/>
    <subsystem name="stabilization" type="quaternion"/>

I watched the video and it looks great. We've done
a lot of research here about VTOL transitioning from hovering to
horizontal flight and vice versa so I'm very interested in your work.
We also like your teams work on al the imav events!!! always impressive to see. So I`m happy your team is active again.
 
Felix said he has fixed issued 132 in the dev branch. How are you
going to tackle horizontal autonomous flight? Would it be as simple as
telling the aircraft to fly with with one of it's axes at a low angle

A lot of the work is done by the quad-shot team. so using the code like it is now you can fly your vertigo in RC stabilised mode. what you do in <define name="MODE_MANUAL" value="AP_MODE_TOYTRONICS_HOVER_FORWARD"/> is controlling the set-point with your RC.
pushing the pitch stick forward will make it fly horizontal, neutral will make it hover.

the cool thing is when hovering: the RC roll stick will activate the vertigo rudder (or differential rpm of the 2 motors), when flying forward: the RC roll stick will activate the vertigo elevons.

or what are the problems you foresee?
constant perfect g turns will let the imu flip over. so the gps compensation is the key here.
this will work nice in forward flight, because the assumption for the gps compensation is that it thinks you are an air-plane: nose in direction of flight.
so the problem will be when to use gps compensation (airplane) to now your heading, en when to use only mag to know your heading (quadrotor) 
when making it autonomous all these problems have to be solved!

an other option can be do not fly perfect g turns ;)
Bart Remes

Best,
Chris

On Wed, Feb 22, 2012 at 3:59 PM, Bart Remes <address@hidden> wrote:
> the status is
> http://www.youtube.com/watch?v=vaxpE3svamo&list=UUGS-Ra1fPexk01KYIBeySRQ&index=2&feature=plcp
>
> still have to implement (and test) quat initialisation. very important when
> your imu is no in a normal oriantation
> https://github.com/paparazzi/paparazzi/issues/132
>
> and the atmov team is doing the best to implement the gps coordinated turn
> for int_cmpl_quat.
> first step is test this:
> https://github.com/paparazzi/paparazzi/issues/130
>
>
> Bart
>
> On Wed, Feb 22, 2012 at 1:08 AM, Chris Wozny <address@hidden> wrote:
>>
>> Does anyone at TU Delft know the status of this project? I'm looking
>> at it and it looks quite robust. I'd be curious to try it with our
>> Mini-Vertigo. Are there any limits to the control laws for this branch
>> like trying to fly a fixed-wing at 90 degrees? The quadshot would see
>> similar issues from autonomous control and singularities as ours does
>> and this does look like the right direction for our project. Thank you
>> for pointing me this out to me Bart and Roman!
>>
>> - Chris
>>
>> On Tue, Feb 21, 2012 at 4:22 PM, Roman Krashanitsa
>> <address@hidden> wrote:
>> > I looked at the
>> >
>> > https://github.com/tudelft-atmos/paparazzi/tree/quadshot/sw/airborne/firmwares/rotorcraft/stabilization
>> > branch and the changes look very solid. I think this is the way to go.
>> >
>> > Roman
>> >
>> >
>> > 2012/2/21 Bart Remes <address@hidden>
>> >>
>> >> chris,
>> >>
>> >> look at the https://github.com/tudelft-atmos/paparazzi/network
>> >>
>> >> they are making the quatshot autonomous.
>> >>
>> >> It will do basically the same as the vertigo but with
>> >> other controls organs.
>> >>
>> >> Bart Remes
>> >>
>> >>
>> >> On Friday, February 17, 2012, Roman Krashanitsa wrote:
>> >>>
>> >>> Chris, it was not very much hovering but flying at 80-85 pitch. Moving
>> >>> very slowly.
>> >>> But then it can transition to essentially 7 deg pitch horizontal
>> >>> flight.
>> >>>
>> >>> Yes, it works totally fine in auto2.
>> >>>
>> >>> Roman
>> >>>
>> >>> 2012/2/17 Chris Wozny <address@hidden>
>> >>>>
>> >>>> Felix,
>> >>>>
>> >>>> I know in the past when we used IR sensors instead of an IMU, we just
>> >>>> mounted the IR sensors on this aircraft such that it thought it was
>> >>>> level when flying at high angles of attack. This is why I believe
>> >>>> such
>> >>>> a pitch offset will allow the aircraft to fly horizontally with high
>> >>>> AoA's (especially since Roman has worked with essentially a prior
>> >>>> version of the same aircraft.) I understand what you're saying about
>> >>>> the mechanics being different when hovering vs. horizontal flight.
>> >>>> However, this aircraft will just fly around horizontally with a very
>> >>>> high AoA and is capable of hovering in AUTO1 (my memory is foggy
>> >>>> about
>> >>>> it hovering in AUTO2.) I hope that my explanation is a little clearer
>> >>>> than mud :)
>> >>>>
>> >>>> - Chris
>> >>>>
>> >>>> On Fri, Feb 17, 2012 at 1:00 PM, Felix Ruess <address@hidden>
>> >>>> wrote:
>> >>>> > Hi Chris,
>> >>>> >
>> >>>> > I'm not sure what you expect by telling the aircraft it is
>> >>>> > horizontal
>> >>>> > while
>> >>>> > it actually is pitched up at 80deg. You can do that, but what
>> >>>> > exactly
>> >>>> > do you
>> >>>> > hope to achieve by that?
>> >>>> > The fixedwing firmware really won't work properly with that
>> >>>> > (without
>> >>>> > modifications that is), as the guidance/stabilization there assumes
>> >>>> > a
>> >>>> > CTOL
>> >>>> > without hover capability. Also stuff like roll basically becomes
>> >>>> > yaw
>> >>>> > when
>> >>>> > you suddenly hover, etc.
>> >>>> > If you have an aircraft capable of both, VTOL/hover and transition
>> >>>> > to
>> >>>> > normal
>> >>>> > flight, you probably want to base that on the rotorcraft firmware
>> >>>> > (just like
>> >>>> > the quadshot, but which is arguably much more like a quadrotor in
>> >>>> > hover) or
>> >>>> > maybe even make a new firmware taking from the others what you
>> >>>> > need...
>> >>>> > What we call a firmware is "just" a main.c with a collection of
>> >>>> > appropriate
>> >>>> > peripherals and subsystems listed in a bunch of makefiles.
>> >>>> > Ok, maybe that is a bit simplified, but basically that's it ;-)
>> >>>> >
>> >>>> > Regarding the singularities (of the euler angle representations):
>> >>>> > Even if your estimation does not have any (e.g. ahrs
>> >>>> > int_cmpl_quat),
>> >>>> > if the
>> >>>> > control is using euler angles, you can of course still run into
>> >>>> > singularities there... basically anything that takes euler angles
>> >>>> > as
>> >>>> > inputs
>> >>>> > is potentially prone to problems around the singularities.
>> >>>> >
>> >>>> > Cheers, Felix
>> >>>> >
>> >>>> > On Fri, Feb 17, 2012 at 8:31 PM, Roman Krashanitsa
>> >>>> > <address@hidden> wrote:
>> >>>> >>
>> >>>> >> Chris,
>> >>>> >>
>> >>>> >> Most possibly, yes. Also this is easy to check - just make
>> >>>> >> modifications
>> >>>> >> and look at the telemetry values, it should show 0 deg pitch at
>> >>>> >> aircraft
>> >>>> >> hovering orientation.
>> >>>> >>
>> >>>> >> Roman
>> >>>> >>
>> >>>> >> 2012/2/17 Chris Wozny <address@hidden>
>> >>>> >>>
>> >>>> >>> All,
>> >>>> >>>
>> >>>> >>> If I have a fixed-wing aircraft capable of hovering at high angle
>> >>>> >>> of
>> >>>> >>> attack (80-90 degrees), would it be possible for me to just
>> >>>> >>> change
>> >>>> >>> the
>> >>>> >>> BODY_TO_IMU rotation for theta to that angle of attack so that it
>> >>>> >>> could
>> >>>> >>> hover at that angle? I'm sure I'd have to adjust some PID gains
>> >>>> >>> for
>> >>>> >>> it to
>> >>>> >>> function properly. Also, if this is possible, would I have to
>> >>>> >>> worry
>> >>>> >>> about
>> >>>> >>> singularities at that angle even though I'm using complementary
>> >>>> >>> quaternions
>> >>>> >>> on the dev branch?
>> >>>> >>>
>> >>>> >>> Best,
>> >>>> >>> Chris


reply via email to

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