Same news for Bremen,
We
have tested a low cost Sparkfun IMU ( RAZZOR
IMU 6DOF Razor - Ultra-Thin IMU)
with the paparazzi hardware.
Five month ago we tested the tiny13 with or quadrocopter code
successfully http://vimeo.com/9894675
. We were a little bit surprised, that we can fly our quadrocopter with
low
cost MEMS sensors.
The price of the sensors is about 9.9 euro and 6.99 euro in single
quantities (lipoly.de). The quality is not as good as the ADXRS610 but
we found
no significant differences in fly with it since the drift is
compensated. For
the quadrocopter we use our
Klaman-filter (run time is 7.5 ms)
and for normal plane we use a filter interdruced by
William
Premerlani,
and Paul Bizard:
Direction Cosine Matrix IMU:Theory, William
Premerlani, Paul Bizard,05.17.2009, http://code.google.com/p/gentlenav/downloads/list
We take this filter for normal plane
configurations and integrate it to the Paparazzi autopilot system by the use of a small planar sensor pub with
new MEMS rate sensors form STMicroeletronics LPR530AL (pitch and roll) the
LY530ALH (yaw) and
the ADX335 (Analog Device) acceleration
sensor. The rate sensors have high vibration frequencies (over 20kHz)
and a
high vibration resistance. After exact bias calibration of the sensors
we found
out that the DCM filter with PI-Controller is very robust by low
computer cost.
The run time is only 0,59ms for the LPC2148 ARM7 micro controller
running at 60
MHz. The only small disadvantage is that the controller after power up
needs
about one minute to stabilize on zero pitch and roll angle
We made automatic flights with a 400mm and
800mm span with delta (up to 17m/s. wind speed). Compared
with the old IR-sensors attitude estimation these
solutions give all weather compatibility and a better performance in
flight
dynamics.
The kinematic equations are
integrated in a notation
of the DCM 50 times per second. To save orthogonality of the DCM for
every
integration step small adjustments of the matrix elements are applied.
For bias
and drift error correction a PI
feedback controller is used. To estimate the errors the
acceleration
data and the GPS ground speed is taken into account.
The
price of the new sensor set is lower than the old
IR system. For the twog the resistors to ground had to be extracted. A low cost system consist of 69$
for the twog (I fetched it for
this price currently 125$) 35$ for
GPS and 90$ for the IMU. If you can solder the
imu self it will cost
less than 35 $. You have to short the
high pass filter on the rate signal for better performance of the
RAZZOR IMU.
Currently a paparazzi compatible planar pcb is under development with this sensors (30mm*57mm) with normal
servo connectors to fly quadrocopters and normal planes.
It will
need a few weeks to document the work since most of our team members
are in
vacation. We will do this soon as
possible. The changes on the code are marginal.
Regards
Heinrich
Warmers_Tiny_IMAV2010.pdf