The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
sensors aren’t very accurate and the angular velocity error is
bigger than the terrestrial velocity rotation. Then for the
gyroscopes calibration we use a rotation plate.
The described method (equations 7, 8) takes into account the
bias, axes misalignment and scale factors. The non
orthogonality of the sensor axes is given by three misalignment
angles: given a fixed x axis, the misalignment of y axis is
obtained by a rotation angle of y axis respect to z axis, and
by the rotation angles of the z axis 0 ZX and 0 zy , respect to x and
y axis.
By using the rotation matrices, and assuming small values for
the above defined rotation angles, we obtain the formulation of
the problem:
gx 1
1
0
o"
gx
gx 2
=
1
0
gy
(9)
gx 3.
-3
gzy
1
_gz_
where gx 1? gx 2 , and gx 3 are the components of the acceleration
along the misaligned axes.
The equation (9) can be rewritten, taking into account both bias
and scale factors; we obtain the equation:
l + s gx
0
0
gx
=
- 9gy*
1 + ^
0
gy
+
b sy
.
1 + -V
_gz_
K
where l gx , Igy, l gz , are the sensor measures, Sg*, Sgy and s^ are the
scale factors and b gx bgy bgz the bias factors along the axes x, y
e z of the reference frame.
By using the abovementioned model, the state equations have
the form:
gx 2 + gy 2 + gz 2 - \g\ 2 = 0 (11)
ax 2 + cay 2 + ox 2 - \cd [ 2 - 0 (12)
for accelerometers and gyroscopes respectively.
The unknowns of these equations are the misalignment
parameters, the scale factors and the bias factors, and can be
solved by using a least squares procedure.
The combined system of (11) and (12) uses parameters properly
weighed (Krakiwsky, 1990) and must be linearized. For writing
the state equations, the precise determination of the sensor axes
is not required. To have a number of equation exceeding the
unknowns number, it is sufficient to position the IMU in
different orientations, as possible distinct and independent each
other.
3. THE STATIC CASE - THE ANALISYS OF THE
ACCELEROMETER RESULTS WITH KALMAN
FILTER
In this section we describe the test performed on an IMU, for
the determination of bias, scale and misalignment factors of the
accelerometers.
First we have analyzed the accelerometers measures of the tri
axis sensors, obtained in the static case, by using a Kalman
filter.
The utilized instrumentation is a MEMS IMU made by Analog
Devices: the ADIS 16350.
The ADIS 16350 is a complete triple axis gyroscope and triple
axis accelerometer inertial sensing system. For the tri-axis
accelerometer the measurement range is ± 10 g, with 14 bit
resolution, the axis non-orthogonality is ± 0,25 degrees at 25°C
and the bias is 0,7 mg at 25 °C, with a 4 mg/°C temperature
coefficient.
In the model of static case, with constant acceleration, we can
write the following state equations:
*, + i =x,+v,M
■ V ,+I =v, +a,M (13)
where:
x is the position, v is the velocity and a is the acceleration; t and
t+1 indicate the time, At is the time rate of acquisition and % is
the model stochastic error.
For the tri-dimensional case, the matrix formulation is:
V. = ®,*, + *, ( 14 >
where x is the unknowns vector with nine components (three of
position, three of velocity and three of acceleration), O is the
state matrix having 9x9 dimension, and e is the model
stochastic error vector with nine components.
The observation equation is:
z, =Hx ( +v, (15)
where z is the vector of the observation of the tri-axial sensor
with three components (the acceleration measures of the
sensor), H is the measurements matrix (3x9) and v is the noise
sensor vector.
The rate acquisition was 1/200 sec.