[Top][All Lists]

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

Re: [Paparazzi-devel] IMU Problem after BungeeTakeoff?

From: Christophe De Wagter
Subject: Re: [Paparazzi-devel] IMU Problem after BungeeTakeoff?
Date: Tue, 17 May 2011 12:42:47 +0200

Dear Heinrich

The the problem is that we "only" use omega x v and not the other mandatory terms in accelerating and rotating reference frames: the bottom equation on is the complete transformation. In our case the sensor does not move with respect to the body axis leaving out vr and ar. Then if your imu is not in the center of gravity you also need omega^2 x imu-location. Then we fill (omega x r) = v-gps which is only correct if the plane has no alpha and beta angles (moves perfectly along x-axis) but you still need the 2 other acceleration terms (which could be simplified to gps-acceleration when alpha and beta are zero). But since they are difficult to measure the are often omitted on purpose in many filters. Although the bungee is short in time, errors grow very rapidly because of the first order linear characteristic of the filter, meaning that large errors even with small filter gains still penetrate in your pitch and roll. With a DCM filter with a 1/50 gain that means that very high g frontal acceleration (specific force points backwards -> almost 90 deg error) every iteration of the DCM 1/50 of 90 degree error grows into your pitch... which can grow to nasty pitch errors when applied during a full second at 100Hz... 

Please also note that during hand-launch with the nose in the wind followed by an aggressive climb this problem is not really visible as the error is usually much smaller than the commanded aggressive climb pitch.


On Tue, May 17, 2011 at 10:47 AM, Prof. Dr.-Ing. Heinrich Warmers <address@hidden> wrote:
i think one of the problem is the fact that  in the cross product of  a=W x V (rates cross speed over ground)  only the
yaw  rates are taken into account to correct the gravitation vector to fly long time curves.
By the way the acceleration time is only  a few seconds. In this time it is impossible to have large errors in the 35° range
generated by the bias of the rate sensors. We had no problems with automatic hand starts.
It may be help, if you wait for about 2 minutes after power on for getting the working point of the feedback control system.
The zero values for the rate must be found very exact (in the range of 1LSB).


antoine drouin schrieb:
same scenario, but this time, i also made the filter without heuristic
use the magnetometer for yaw only ( I believe that is what arduimu
does ).
You notice that the error is even worse ( up to 35 degres now).

On Mon, May 16, 2011 at 3:55 PM, Gautier Hattenberger
<address@hidden> wrote:

The arduimu_basic code has some special tricks to try to detect high
acceleration when launching (high throttle, low speed) and prevents the imu
from using accelerometers. I'm still removing some bugs, but it seems to
work (although it's only a stupid hack in the end...). Antoine's filter
seems to be a bit clever than that.


On 16/05/2011 13:35, Lange Dennis (langede0) wrote:

We're using the arduino-basic code and the Kalman Filter for GPS-Data


Am 14.05.2011 05:15, schrieb antoine drouin:

Hi Dennis

What filter are you using ?

On Sat, May 14, 2011 at 12:38 AM, Lange Dennis (langede0)
<address@hidden>  wrote:

Hi all

We would like to develop an autonomous Bungee-Takeoff. We have written a
routine that actually works. In the first seconds of the start it looks very
good. But then the UAV starts to decline again. But the log file sayd that
the UAV wants to climb all time. And we also saw that the PFD did not show
the right position.
So we believe that the IMU has an error because of the acceleration during
the takeoff.

We also tried the BungeeTakeoff from the OSAM-Team>>  same problem

Has anyone else had this problem? Or does someone have a solution?
We considered to control the angle of attack with an extra sensor during
takeoff. Can it work?

Thanks in advance


Paparazzi-devel mailing list

Paparazzi-devel mailing list

Paparazzi-devel mailing list
Paparazzi-devel mailing list

_______________________________________________ Paparazzi-devel mailing list address@hidden

Paparazzi-devel mailing list

reply via email to

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