GPS/MEMS-IMU Integration
Jieying Zhang, Ezzaldeen Edwan, lunchuan Zhou, Wennan Chai, Otmar Loffeld
Center for Sensorsystems (ZESS), University of Siegen
57068 Siegen, Germany
Abstract- The integration of inertial sensors is a widely-adopted system under rapid development, thus keeping the whole
method for covering GPS outages. However, the position system in an affordable manner. With more sensors
accuracy in vertical direction is often challenged in comparison incorporated, more information is fused in the positioning
with the one in the horizontal plane. This paper explores the algorithm, which leads to meet the ever-increasing demand in
possibility of using MEMS barometer as an aug mentation in the navigation accuracy, robustness etc. The barometer in the
GPS/INS integration in land application. A prototype system, multi-sensor system can improve the availability and accuracy
composed of a commercial GPS receiver with low-cost MEMS
of the positioning solution in the vertical direction. While
IMU and barometer on a DSP platform, is implemented and
similar to other MEMS sensors, the critical point of integrating
described. In addition, the initial investigation of the MEMS
the barometer is also to deal with the limited sensor
barometer is addressed. With the field experiment data collected
performance, since the ambient air pressure measured are
from a car-driving test in a hilly region, the benefit of barometer
influenced by fluctuations in ambient temperature, ventilation
aiding is demonstrated.
and so on. Recent researches have reported the successful
Keywords; formatting; MEMS barometer; GPSIINS application of the barometer in indoor or personal navigation,
for example to identify different building floors. Some of the
researches also consider a second barometer as a reference. [2]
I. INTRODUCTION
This work explores possibility of applying MEMS barometer to
Since decades, the integration of the Global Positioning the land vehicle navigation and further investigates the system
System (GPS) and the Inertial Navigation System (INS) has performance of the multi-sensor integrated system in the
received fast development due to the fact that it delivers a vertical direction. Considering the practical setup of the
higher-quality navigation performance, not only in terms of system, only one barometer is considered in this work.
accuracy but also system robustness, by taking advantage of
With the purpose of implementing a consumer navigation
the individual superiorities of each standalone system. A
system for embedded applications, the development of a low
widely applied field is the land vehicle navigation, where the
cost multi-sensor integrated navigation system based on a
GPS outages are expected to be bridged in the GPS challenged
Digital Signal Processing (DSP) platform is implemented and
environments like the urban canyon or foliaged areas.
addressed in this paper, with the emphasis on the system
It is observed that the position accuracy is more challenged development of the barometer augmentation. While equipped
in the vertical direction than the horizontal dimension, even in with a MEMS barometric altimeter apart from the MEMS
the GPSIINS integrated system. Normally, the altitude IMU, the prototype system is expected to behave more robust
accuracy is considered critical in the aerial applications, while and demonstrate the superiority in vertical performance.
appears less important for the land vehicles. However, with the
The remaining of the paper is organized as follows: In the
ever-increasing demand in navigation accuracy and robustness,
second section, the fundamental of barometric height is
for example to know the exact level in the cloverleaf junction
introduced, as well as its relationship with GPS height and the
or to track a vehicle in the hilly area, vertical accuracy bears an
characteristic of the sensor errors. Moreover, the initial
increasing weight in the land applications.
investigation of MEMS barometer is carried out which
For a long time, pressure sensors like barometric altimeters facilitates the implementation of the multi-sensor integrated
are special-purpose devices which derive altitude from pressure system. The realization of the prototype is described in the
measurements. Yet, considering barometers to enhance the third section, including the system configuration and the sensor
system performance in the vertical direction is not a fresh new fusion and barometer augmentation mechanism. The field
idea. [1] Nevertheless, the cost-driven factor and the physical experiment with an auto is described in the fourth section. The
dimension have been the major barrier of bringing such test results is demonstrated and analyzed in the vertical
integrated systems to the consumer market until the emergence direction. Finally, conclusions and further steps are addressed
of the Micro Electro Mechanical Systems (MEMS) technology in the fifth section.
recently. Benefited from the low-cost and compact MEMS
sensors available, the applications of MEMS barometers are
going widely into the mobile and personal devices and
consequently promoting the multi-sensor integrated navigation
It's well-known that the barometer measures the absolute Indicated from the bottom plot of Fig. 1 , the main error
ambient pressure. The measured pressure P can simply be source of the barometric height is composed of the bias term,
drift term and noise term. The bias term, also known as the
converted into height Hbaro using equation (1 ), where Po is the initial sensor bias, is constant and can be removed by
absolute pressure at the sea level. calibration. The drift, which is caused by the slow change in
[ I1
ambient pressure and other uncertain factors like temperature,
2000
The GPS height in the frequently-used LLH (Latitude,
Longitude and Height) format often refers to the height above 1500
ellipsoid, which is described by the standard earth coordinate
1000
frame of WSG84 for example. The relationship between the
barometric height and GPS height can be expressed in equation 500
(2). 0
217 218 219 220 221 222
h=H+N (2) Barometric height [m]
where h denotes the ellipsoid height, e.g. GPS height; H Figure 2. Histogram plot of barometric height
l::::
modes.
oS 360
E As the barometer needs certain time for the conversion of
0>
340
the pressure measurement, tradeoff needs to be sought between
320 the sensor oversampling rate and noise reduction when it is
0 200 400 600 800 1000 1200 1400 1600 1800
Time [s1 integrated in the system. Therefore, the sensor performance
under different sensor modes is investigated and demonstrated
in Fig. 3. The standard deviation of each sensor oversampling
mode is given in Table 1 , where the maximum conversion time
of each mode is also listed. As expected, with longer
conversion time, the noise level of barometric height is
600 800 1 000 1200 1400 1600
reduced. Nevertheless, the ultra low power mode is still of
Time [s1 great interest for two reasons. First, as indicated in Table 1 , the
difference of the standard deviation in height between the ultra
Figure I. GPS height vs. barometric height in static position high resolution mode and the ultra low power mode is
relatively moderate. In this case, the ultra low power mode
In Fig. 1 , the GPS height varies up to 40 meters in a short shows an increase of several tens of decimeter in standard
time due to the poor observations as well as multipath effects. deviation. Second, with less conversion time, the height
The barometric height converted from the pressure solution from barometer can be available at higher frequency,
measurements using equation (1 ) differs largely from the GPS which may benefit the sensor integration.
599
Moreover, although the measurements under different illustrated in the bottom diagram of Fig. 4. However, as
modes were taken at the identical position and approximately summarized in Table 2, the standard deviations of each test
the same time, the corresponding absolute heights derived show a similar noise level of the barometric heights during
show a slight variance in the top plot of Fig. 3. The disaccord different times.
presents a certain relationship with the ambient temperature
measured as shown in the bottom plot of Fig. 3.
TABLE IT. STANDARD DEVIATION OF BAROMETRIC HEIGHT IN FIG. 4
t:+--- d
-
sensor oversampling modes. Each row indicates the standard
-
deviation of barometric height obtained at the same time but
under different mode, whereas each line shows the noise level
22 51- - ----::5':-,0 --------10-,--' 0--------
0 -- . . ---
-
15-,--
---
-' 0 -----2:-'-0,-
0 -----2:-':-50,------,
300 ,-' under the same mode while at different times. Table 3
Time[s]
demonstrates again that under the same mode, the noise levels
of the barometric heights during different time are similar and
the noise level is reduced with longer conversion time (higher
Figure 3. Barometer reading in static position with different modes
sensor oversampling mode).
TABLE I. STANDARD DEVIATION OF BAROMETRIC HEIGHT IN FIG. 3 TABLE Ill. SUMMARY OF STANDARD DEVIATION OF BAROMETRIC HEIGHT
Height std.(m) 0.47 0.48 0.56 0.62 Std.(m) @ Time 3 0.683 0.580 0.485 0.424
The barometric heights of an identical position recorded Std.(m) @ Time 5 0.637 0.562 0.480 0.535
with the same oversampling mode (ultra low power mode) Std.(m) @ Time 6 0.636 0.601 0.430 0.385
during the successive days are shown with different colors in Std.(m) @ Time 7 0.609 0.639 0.494 0.496
Fig. 4.
Std.(m) @ Time 8 0.725 0.629 0.566 0.484
440
An additional test is conducted to observe the sensitivity of
F,.""'_""_="""'
, .""
,,,-""
,,
""
.,,,""
,,=:=""
__ ._.t=
""== _ =;;.
...
.
""
"=
=. ,.. _= .. . . _.=
" _
-."' _.. ,==.,,.
... .... the sensor's reaction upon height changes. The barometer is
j 400 -
'''' .. .... ... . .. .. ... -"'
. = . ...
. _ . "'- .. ... ....
.... .
.-__________________________
______
moved up and down by hand at approximately 0.5 Hz with a
380 height difference of l. 5 to 2 meters. The result is illustrated in
3800---------,00O---1---2O- Fig. 5. It is shown that the waving of hand up and down is
Time Is] captured by the sensor, though a small drift of the slope is
24,------,------,------,,- -----,------,------,
. ;." : : . -.j
observed due to the ventilation change during the test.
F6
Barometric height
224
1.50\-----,;;- ------;;';
O
l O ,---
;
:
----,;,';;-----
-- "fn- ------:-2"'" ,--------;;(
Time Is]
223
I 222
221
] 220
219
Figure 4. Barometric height in the same position during different days
218 '
' 2-----:'':-4 ------, 7'-' 6 ------=--
':":, 0,..----:'::
8-" --::':02 -----:2-:-:2-----=2':-4 --
The absolute heights in different time periods vary each Time [s]
time due to the different weather conditions, where the ambient
temperature measured again indicates one of the influences as
600
Figure 5. Sensitivity test of barometer B. Sensor Fusion and Barometer Aiding
-2 ud"--,_tV"::/A':.s'
Estimated ' n
The prototype system integrates one commercial GPS ... ----------. r--J---,!!i' ttit-"'" ______
....r-----
...:!++(E+ r--
I
I
receiver, one low-cost IMU and one MEMS barometric :I angular
Tri-axis. ro bl:_
---- -, + }_+j ,'"" , o, t). i'7NS
E ln ' t , d
velocity : + Strapdown + + ....: Navigation
altimeter on a DSP platform, taking advantage of its powerful I: . .
-b :
INS Estitted rl'S :I output
+ +
L _co:c J
Tn-aXIS positIon
processing ability. The block diagram of the prototype is +
L __ j-=""'--'I---'-r-It-t
shown in Fig. 6.
t::ec
Estim ted
ac cr
sib
gyro blils
Swb
-------------------,
OPS
bws
rops +
I
1 E!::S
posllton
I +
OPS
:
I u- uu_ _
velocity
V GPS
GPS data stream I I
I+----,.-I UART
:
GPS
---
:measurement
decoding
---I
I I
Navigation
Solutions I Estimated
attitude
Time pulse measurements
I
I PS G
r hGPS POOP,
B
I
h g t
L...VDO,
....l! P
I--------j
I Barometer
1 I
I
I
I
:I Barometrie_
height hbar'1 LPF Bias
L ________
I I Smoothed Calibration
I
I
i
b.1ro
l netrie
. Corrected
Barometric
I he1ght
height bias barometric
I
_
+ height
_____1 +
601
GPS antenna. The calculated DOP, which indicates the satellite
geometry, is currently applied as the criteria of choosing a
good-quality GPS solution. The barometer altitude is calibrated
by the estimated bias term. Then, the calibrated barometric
altitudes are used to constrain the GPS height and provide an
additional vertical observable to the least square estimator
when the GPS is under challenged environment.
Figure 10. Elevation profile in Google Earth, maxim height difference 0[80
meters
B. Result Analysis
The low-cost IMU was mounted on the top of the car roof,
as well as the GPS antenna which was about half meter ahead
to avoid the interference caused by the strong magnetic field.
The processing board of the prototype system, which is shown 100
seat and was about half meter under the car roof. A black
plastic housing was made for the barometer. Though not
sealed, it shaded the sensor against strong light.
The trajectory was chosen to drive between two hills in the
city of Siegen, Gennany, as depicted in Fig. 9. The total
duration of the car-driving was 1 5 minutes. The altitude
difference in this route is about 80 meters and is illustrated with '
X10
Time of the week [s) )(10'
the elevation profile from Google Earth in Fig. 10. Moreover,
this route also includes a part of highway and a part of Figure I I. GPS height vs. barometric height
residential area.
602
The height results are in accord with each other with the
same slope. The top plot in Fig. 11 shows again the 50.912 2D position
350
-;- 150
O'h
Jj 100
1
50
L---;3:-:.1""72--------;
-: 3.1"'7;C =-
4 -----;3-:0;- 1. 7;;:6----;3c-:."'71 8 ,-------J ;17053.1713.17153.1723.1725 250
Time [s] 105
Time [5] 105
x x
200
L ::-:
3. 17 1----3-: :-.1':::c7=--
2 -=-3.17:-::- :-- 3-=-.-:':
3-----= 3.1-=-74 17-=-: 5 ---- =--=-- 3.17
3.-=- 17::-:6----3-: :-.1':::c
77 8----3
:-::- -= .1=-::-'
79
Figure 12. Smoothed barometric height 5
X 10
lime of the week [s]
-R - j ,
i 300 -- -- I _ - I - --: , -
1"'...... I ........ I
--- L_....--l' -- I ' .... ..J First, the roof window of the car was half open to let the
__ -,
250 - wires go through during the experiment. Therefore, certain
-- ,
1 amounts of ventilation were observed during the car drive.
200
50.92 ......
_ ..... However, its minor influence can be neglected in this
- 8.055
50.9t 8.05
8.045
experiment.
50.9 8.04
Latitude 8.035
8.03 Second, the measured ambient temperature observed a
50.89 8.025 Longitude
large change as depicted in Fig. 1 6. Nevertheless, the
barometric height was not much impacted by the temperature
Figure 13. 3D integrated position change.
603
REFERENCES
Ambient temperature during car drive
[1] N. Hayashi, "Augmentation of GPS with a barometer and a heading rate
gyroscope for urban vehicular navigation," Department of geomatics
engineering, M.S. dissertation, University of Calgary, 1996.
[3] "BMP085 digital pressure sensor data sheet ," Bosch Sensortec, 2009.
19 'c:.,7c-
3-"= , ----:
.
3-'=: '7=-2 ----:
3.-'=
: '7=-3 ----:C
3."= '7C--
4 -:C3."=
'7=-S ----;c
3."= ,76=---;:-"::=---;:-":::--
[4] 1. Zhou, S. Knedlik, E. Edwan, O. Loffeld, "Low-cost TNS/GPS with
Time of the \\eek [s]
nonlinear filtering methods," in Fusion 2010 (13th International
Conference on Information Fusion). Edinburgh, UK. 2010.
Figure 16. Ambient temperature during car drive
ACKNOWLEDGMENT
604