Você está na página 1de 7

Advanced Materials Research Vols 383-390 (2012) pp 2060-2065

(2012) Trans Tech Publications, Switzerland


doi:10.4028/www.scientific.net/AMR.383-390.2060

Online: 2011-11-22

The Fine System-level Calibration Technique of Size Effect Error for the
Fiber-SINS
Zhu Lufeng and Zhang Chunxi
School of Instrument Science and Opto-electronics Engineering, Beijing University of Aeronautics
and Astronautics, Beijing, China
mailzlf@yahoo.com.cn
Keywords: The size effect error; Fine calibration; Fiber- SINS; Kalman filter; Error model

Abstract. A fine system-level calibration technique of size effect error is presented for Fiber-strapdown inertial navigation systems (Fiber-SINS) in this paper. When the strap-down system is
mounted onto high precision turntable and rotating along predefined trajectory, the velocity errors
can contain information related to the size effect error, which can be estimated accurately by
Kalman filter. In this paper, principles of the calibration technique are emphasized, the model of the
size effect error is presented and Hardware-in-the-Loop Simulation platform is designed to validate
the efficiency of this technique. Results show that the calibration technique has the advantages such
as simplicity and high precision. Thus it is of great importance for practical applications.
Introduction
The inertial sensors (gyroscope and accelerometer) are mounted onto the body directly in FiberSINS. Ideally, the accelerometers should be mounted precisely at the same position. Obviously, this
is impossible because the accelerometers are of finite size and the design on the positioning of
hardware is limited. Since the accelerometers have physical displacements with respect to the ideal
position, then the tangential and centripetal force sensed by the accelerometers are referred to as
size effect [1].
Provided the resolution of the measurements into the navigation reference frame is implemented
accurately in the presence of continuous rotations, the effect from additional acceleration will
integrate to zero over a few cycles of rotation. But the size effect error is unavoidable because of the
imperfections in the strap down computational algorithms. Of particular concern here is the effect of
oscillatory motions which will be rectified to give steady acceleration errors [1]. Consequently, it is
necessary to apart and extract the size effect errors by fine calibration technique, then eliminate them
by compensatory algorithm.
The FOG can be ignored when concerning the size effect errors because the acceleration does not
affect them. Traditional calibrations for Fiber-SINS generally get raw data by multi-position roll
method and then obtain errors by analysis. But these errors cannot be precisely separated because of
observing constraints. Literature [2] had presented a continuous automatic calibration technology of
accelerometers, but had not given the detailed implementation process of reckoning the size effect
error by system-level technique. Literature [3] had presented a system-level technique for INS, but
had not mentioned the calibration technique for size effect error of accelerometers.
Size Effect Errors of Fiber-SINS
Summarize of Size Effect Errors. Assume that the sensitive axis parallel to the body axis, rp is
the vector that centroid P of accelerometer relative to the origin of body frame and rpx rpx rpx is
its component in body frame which is rotated with respect to the inertial frame by the angular
velocity and with origin acceleration a0 , and a p is the acceleration of the input axis of
accelerometer
a p = a0 +  rp + ( rp )
(1)
All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications, www.ttp.net. (ID: 152.14.136.77, NCSU North Carolina State University, Raleigh, USA-28/04/15,02:13:49)

Advanced Materials Research Vols. 383-390

2061

Where  rp + ( rp ) is the detrimental acceleration caused by size effect, the first part is
tangential acceleration and the second part is centripetal acceleration. Assume that the angular
acceleration in the calibration is extremely slight that  0 , then
aep = ( rp )
(2)
Then analysis the size effect error with the accelerometer of X-axis as an example.
if

rpx = rpxx

rpxz

rpxy

(3)

where rpx = rpxx rpxy rpxz is the component relative to the three axis of body frame.
The effect of y on aepx is
a = -
y
epx

2
y

(r

x2
px

)i

z2 2
px

+r

rpxx

( rpxx 2 +r

z2 2
px

=y2irpxx

(4)

As before, the effect of z on aepx is


z
aepx
= - z2 irpxx

(5)

Since the x parallel to aepx , the size effect error will integrate to zero. Then, the components

rpxz and rpxy can be ignored when calculating the size effect error of accelerometer of X-axis.

rpx = rpxx 0 0
Substitute (4) (5) into (2)
aepx = - rpxx ( y2 + z2 )

(6)

(7)

Similarly,
aepy = - rpyy (x2 + z2 )

(8)

aepz = - rpzz ( x2 + y2 )

(9)

The INS Error Models. The navigation frame used in this paper is a geographic north-east-down
(N-E-D) frame ( oxyz ). Under stationary base, neglected the initial system errors and simplified the
error modal of INS to [3]:


x
y

v x = f x + g y
v y = f y g x
v z = f z
= z y y z + x
= x z z x + y
z = z

(10)

Where g is the local acceleration of gravity, is earth rate, v is velocity error, is attitude error,
f is accelerometer error, is gyro error, and the subscripts are component in navigation frame.
The sensor error models. The sensor errors to be calibrate are constant errors of accelerometer and
gyro: bias, scale-factor, misalignment and size effect. The following model define these errors in
body frame obxbybzb

b = gBb + ( gMAb + I i gSF b )i b

f = aB + (aMA + I iaSF )i( f + aep


b

where

(11)
b

(12)

2062

Manufacturing Science and Technology, ICMST2011

xb gbxb gSFx
b b
y = gby + gMAyx
b gbb gMA
zx
z z

gMAxz xb

gMAyz yb
gSFz zb

gMAxy
gSFy
gMAzy

(13)

fxb aBx aSFx aMAxy aMAxz fxb + aepx


b 2

b
fy = aBy + aMAyx aSFy aMAyz fy + aepy
fzb aB2 aMAzx aMAzy aSFz fzb + aepz
z

(14)
Where gB and aB are the bias errors, gMA and aMA are scale-factor errors, gSF and aSF are
misalignment errors, and aep is size effect error.
The Relationship Between System Errors and Sensor Errors
The change rate of horizontal velocity errors components between the period t = 0 ~ T , are given as
vx (T ) vx (0) = f x + g y
(15)



v
(
T
)
v
(0)
f
g
y
y
y
x

where
f = Cbn (T ) f b (T ) Cbn (0) f b (0)

(16)

= Cbn (t ) b dt

(17)

Cbn is

the transformation matrix from body frame to navigation frame.


Substituting Eq.(12) to Eq.(16) yields
f = Cbn (T ) ( aBb + (aMAb + I iaSF b )i( f b (T ) + aepb (T ) ) )

(
= C (t) ( gB + (gMA + I igSF )i ) dt

C ( 0) aB + (aMA + I iaSF )i( f ( 0) + aep ( 0) )


n
b

n
b

(18)
(19)

Higher derivatives of Eq. (15), (16), (18) and (19) can determine the relationship between the
velocity errors and sensor errors.

Fine Calibration Designing


The Fine Calibration Principle. The fine calibration obtains sensor errors by observing the system
errors in navigation process. The position and velocity errors are the changes of position and
velocity calculated in navigation under stationary base. The system outputs contain information of
sensor errors which contributed to them in the navigation. So the sensor errors can be obtained by
designing rotation sequences, forcing sensor errors separately and observing system outputs.
Rotation Sequences Designing for Size Effect Errors. According to Eq. (15), (16), (17) and
(18), attitude and accelerometer errors contributed to horizontal velocity errors. The attitude errors
which caused by gyro errors can be eliminated by calibration and compensation. The accelerometer
errors only related to the system starting and ending position, which means only the size effect error
contributed to the accelerometer errors when system rotated with different angular velocity but
same starting and ending position. Then the size effect errors can be estimated by observing the
change of horizontal velocity while the system is rotating at a certain position.

z (t )

Figure 1.

x (t )

Rotation Sets

Advanced Materials Research Vols. 383-390

2063

1 0 0
0
Cbn ( 0 ) = Cbn (T ) = 0 1 0 , f b (0) = f b (T ) = 0
0 0 1
g
1 0 0
0
Rotation 2 n
Cb ( 0 ) = Cbn (T ) = 0 0 1 , f b (0) = f b (T ) = g
0 1 0
0
Rotation 1

Cbn ( 0 ) and Cbn (T ) are transformation matrix from body frame to navigation frame of the two

observation time with same system position, and the calibration started when t=0. Then z (0) = 0
in rotation 1 and x (0) = 0 in Rotation 2
Observation Equation of Size Effect Errors. The observation equation of size effect errors
can be obtained in the rotation sequences above. In Rotation 1
f = Cbn (T ) ( aB b + (aMAb + I iaSF b )i( f b (T ) + aep b (T ) ) )
(20)
Cbn ( 0 ) ( aB b + ( aMAb + I iaSF b )i( f b ( 0 ) + aep b ( 0 ) ) )
= ( aMAb + I iaSF b ) ( aep b (T ) aep b ( 0 ) )
Expand Eq. (20) and ignore second order terms, then
aepx (T ) aepx ( 0 )
z2 ( T ) rpx

f = aepy (T ) aepy ( 0 ) = z2 (T ) rpy

aepz (T ) aepz ( 0 )
0

gBx gSFx

C (t ) gBy + gMAyx
gB gMA
zx
z
n
b

gMAxy
gSFy
gMAzy

(21)

gMAxz 0

gMAyz 0 dt
gSFz zb

(22)

Use the residuum error of FOG after compensation, and the attitude error can be ignored in a
short time, then Eq. (15) becomes
vx (T ) vx (0) = f xn = z2 (T ) rpx
(23)

v y (T ) v y (0) = f yn = z2 (T ) rpy

(24)

Similarly, in rotation 2
v y (T ) v y (0) = f yn = x2 (T ) rpz
Higher derivatives of Eq. (23), (24) and (25) can determine the equation of error observation

(25)

Table I.

Equation of size effect errors

In Rotation 1

v (T ) vx (0)
rpx = x 2

z (T )

v y (T ) v y (0)
rpy =

z2 (T )

In Rotation 2

rpz =

v y (T ) v y (0)
x2 (T )

Kalman Filter Designing. The system theoretical velocity in fine calibration process is zero
because there is no line movement. In this condition, the designing of Kalman filter could use zero
velocity as external observation and then keep calculating the change of system velocity.
The Filter equation is
v x 0


d v y 0
=
dt vx 0


vy 0

W is system noise.

0 1 0 vx

0 0 1 v y
+W
0 0 0 vx

0 0 0 vy

(26)

2064

Manufacturing Science and Technology, ICMST2011

The measure equation is


Z = [1 1 0 0] X + V

(27)
T

Where X = vx v y vx v y , and V is measure noise. Estimating the state variable by


Kalman filter algorithm, measure updating cycle is 1s.
Hardware in the Loop Simulation Test
Hardware in the Loop Simulation Testing Designing. A simulation test for size effect errors
calibration was designed in this paper to verify the validity of this calibration. Mounted the IMU
onto the platform, collected and calculated the data after the platform had been started up, then
calculated the size effect errors by observing the system velocity errors. The characteristics of
sensor errors of the IMU used are as table 2.
Table II.

Characteristics of sensor errors

Accelerometer
Valve
Unit
500
g

Sensor Error
Bias
Scale Factor
Error
Misalignment

500

ppm

Unit
/h

200

ppm

( h) Hz
After compared the data before and after the compensation of size effect errors, the validity of
this calibration can be verified.
Test Conditions. In the calibration process, the rotation mode of system was as Fig.1of 2.2, the
rotary speed was as Fig.2, and the initial values of Kalman filter in calibration are
2
2
2
2
2
P ( 0 ) = diag {( 0.01 m s ) , ( 0.01 m s ) , ( 0.1 m s ) , ( 0.1 m s 2 ) ;
Q ( 0 ) = diag {( 0.02 m s ) , ( 0.02 m s ) , ( 0.04 m s ) , ( 0.04 m s ) } ;
100

Noise

R = diag ( 0.001 m s )

Gyroscope
Valve
0.5

50

Hz

100

0.05

};

2 2

2 2

3 2

The test of proving the effect of size effect error compensation was as Fig.3, kept the rotation axis
upright and observed the output of the horizontal acceleration:
rate of Z deg/s
rate of X deg/s

Route Set I
600

600

400
300

Angle Rate deg/s

400
200
0

10

15

20

25

30

35

Time s
Route Set II
400

100
0
-100
-200
-300

200
0

200

-400

10

15

20

25

30

35

50

100

150

200

250

300

350

400

Time s

Time s

Figure 3. Verification of Rotation Rate


Figure 2. Schematic Diagrams of the
Sequences
Analysis of the Test Results. After the test above, the calibration results of size effect errors are
as Table 3
Table III. Calibration test results

Size Effect Errors (m)

Axis X

Axis Y

Axis Z

0.038

0.043

0.051

The acceleration outputs before size effect errors compensation in the test were as Fig 4, and the
acceleration outputs after compensation were as Fig 5.

100

200

300

400

-1
-1.5

100

0
-0.5
-1

100

200

Time -s

300

400

300

400

0
-200
-400

200

400

0.01

-0.01

-0.02

Time -s

200

400

Time -s

0
2

0.5

-1.5

200

200

Time -s
AccelerationZ -m/s

AccelerationY -m/s

Time -s

-0.5

-1

-1.5

100

200

300

400

Time -s

Figure 4. Acceleration Outputs before


Compensation

AccelerationY -m/s

-400

AccelerationX -m/s

-200

-0.5

400

2065

AccelerationZ -m/s

200

0.5

Angle Rate -deg/s

400

AccelerationX -m/s

Angle Rate -deg/s

Advanced Materials Research Vols. 383-390

-0.01
-0.02
-0.03
-0.04

200

Time -s

400

0.15
0.1
0.05
0
-0.05

200

400

Time -s

Figure 5. Acceleration Outputs After


Compensation

According to Fig.4, the acceleration errors caused by size effect increased dramatically with the
angle rate before compensation; according to Fig.5, the effect of angle rate to the acceleration
output can be ignored after the size effect compensation. This verified the fine calibration
technique.
Conclusion
This paper presented a fine calibration technique of size effect errors for Fiber-SINS, designed a
series of rotation sequences, and calculated the size effect errors of accelerometers by observing the
system output. The fine calibration technique is convenient, automatic and effective verified by
hardware in the loop simulation test.
References
[1] Zhang Kaidong, The continuous automatic calibration technology of laser gyro strapdown
inertial navigation system (LSINS),Changsha Graduate School of National University of
Defense, 2002
[2] Zhang Tianguang .et. Strapdown Inertial Navigation Technololgy (2nd Edition) ,Beijing
National Defense Industry Press 2007
[3] Robert M.Rogers. Applied mathematics in integrated navigation systems,AIAA, 173-183,
America: American Institude of Aeronautics and Astronautics, 2003
[4] Chen Zhe. The principle of strapdown inertial system,Beijing: Space Navigation Press, 1986
[5] Gu Hongqiang, Yuan Yaxiong, Bai Weibing. Missile Strap-down Inertial Measuring Unit Fasr
Postion Demarcation, 200731(6)719-721
[6] Mohinder S.Grewal. Application of Kalman Filtering to the calibration and alignment of
inertial navigation systems. IEEE, 1991, VOL.36, NO. 1:4-12
[7] Eduardo Nebot, Hugh Durrant-Whyte. Inertial Calibration and Alignment of Low Cost Inertial
Navigation Units, Journal of Robotics Systems,1999,V1(16):81-92
[8] Xu Bo, Sun Feng. A FOG Online Calibration Research Based on High-precsion three-axis
Turntable, 2009 International Asia Conference on Infornatics in Control, Automation and
Robotics.
[9] Huang Ping, Research on Calibration Technique of the Strapdown Inetial Navigation System,
Harbin, Harbin Engineering University master thesis, 2005.

Manufacturing Science and Technology, ICMST2011


10.4028/www.scientific.net/AMR.383-390

The Fine System-Level Calibration Technique of Size Effect Error for the Fiber-SINS
10.4028/www.scientific.net/AMR.383-390.2060