[Top][All Lists]

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

Re: [Paparazzi-devel] IMU's

From: Christophe De Wagter
Subject: Re: [Paparazzi-devel] IMU's
Date: Thu, 1 Sep 2011 20:17:55 +0200

Hear Hector,

the DCM_float implementation in paparazzi indeed only estimates attitude (called AHRS in paparazzi) but it does use GPS velocity to compensate centripetal forces and to estimate heading (magnetometer could be used)


-assumes all velocity is in body X direction (= aircraft model)
-does not compensate longitudinal acceleration (change is velocity)
-does not improve GPS measurements using accelerometers (called INS in paparazzi)

On Thursday, September 1, 2011, Hector Garcia de Marina wrote:
Hello professor,

I have writen about the DCM algorithm (mahony papers) only estimates the attitude, not the 3D position as it does not take into account any position observation, as GPS, vision, or whatever :P. But of course, the DCM algorithm estimates the attitude, althougt as it only relies in accelerometer for measuring the gravity, the algotihm can not estimate the yaw angle.


On Tue, Aug 9, 2011 at 11:17 AM, Prof. Dr.-Ing. Heinrich Warmers <address@hidden> wrote:
you're are right that in  the current implementation only compensate the centrifugal  force  for  circles  around  the z axis.
It it not a problem  to solve   vector f =  vector omega  (vector multiplication ) vector V.
By adding this simple equation also to  the AHS of paparazzi  it would be possible to fly acrobatic with normal planes for a long time period.
I think you're  are wrong,  that the current implementation  oft  the DCM  do not estimate the 3D attitude.
We have flown circles for 20 minutes  (angel up to 70° radius down to 70m).
We have also flown in clouds and circles up to to 1700m  ground level without  problems.
The paparazzi estimator (estimator.c) calculate position and speed and fill the gap between  the 250ms until the next  gps information.
It would be a nice work to implement  acrobatic flight components like  F3A  in Paparazzi.


Hector Garcia de Marina schrieb:

Hello Todd, as the DCM algorithm does not estimate the 3D position/velocity, it can not compensante the centripetal acceleration. You can still using the DCM algorithm but with poor performance, specially when you have mid/large wz (gyro in z axis) for instance in a turn, even at low velocities.

Note that even if you have a GPS, the ground speed will have a delay (depending on your GPS could be even more than 1 sec). So if you do not take account this, for non-stationary flight (under linear-accelerations) the performance of the DCM algorithm stills poor. As always, the quality performance depends on your mission requirements.


On Tue, Aug 9, 2011 at 2:30 AM, Todd Sandercock <address@hidden> wrote:
From just my reading into the Paparazzi's DCM code, it also does centripetal acceleration compensation via GPS... as long  as "USE_GPS" is defined.


From: Prof. Dr.-Ing. Heinrich Warmers <address@hidden>
To: address@hidden
Sent: Tuesday, 9 August 2011 1:31 AM
Subject: Re: [Paparazzi-devel] IMU's

Hi Oli,
I think the rating  of the IMU`s depend of there vibration resistance and the bias of the rate sensors.
I think there is the following rating:

1. Bosch (SMG074)   2. Analog device  (ADXRS620..)  3.Melexsis  4.  Invensense  5. ST electronic

The Bosch and the Analoag device  are made for automotive applications and therefore they deliver a maximum of vibration tolerance.
For normal planes the best algorithm in my opinion is the  dydrohnes DCM, since this  can correct  the attitude given by integration of the rates
even when continuos circles are flown.
You have to consider the centripetal forces. Every IMU without measuring the speed over ground can't do this.

If you want to test the Analog device types  you can fit sparkefun modules (  ADXRS613-Breakout  )

 to a box (total cost < 80 Euro). This is software compatible to the paparazzi DCM implementation.  You can find them  in the repository of Christian Niemann.
I think this will also solve you problem with IC-motors.



reply via email to

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