W.D. Ding, J.L. Wang, P. Mumford, Y. Li, C. Rizos. Time Synchronization Design for Integrated Positioning and Georeferencing Systems
Proceedings of SSC 2005 Spatial Intelligence, Innovation and Praxis: The national biennial Conference of the Spatial Sciences Institute, September,
2005. Melbourne: Spatial Sciences Institute. ISBN 0-9581366-2-9
TIME SYNCHRONIZATION DESIGN FOR INTEGRATED
POSITIONING AND GEOREFERENCING SYSTEMS
W.D. Ding, J.L. Wang, P. Mumford, Y. Li, and C. Rizos
School of Surveying and SIS, the University of New South Wales, Sydney, Austrlia
Weidong.ding@student.unsw.edu.au
Key words: GPS, INS, integration, time synchronization
ABSTRACT
More and more surveying companies are embracing digital technologies such as CCD cameras and LiDAR for aerial
photogrammetric mapping and imaging applications. In this trend, direct georeferencing plays a crucial role. The
obvious benefits of direct georeferencing are reductions in the requirements for aerial triangulation and ground control
points, both of which account for a significant cost in survey work. In the case of direct georeferencing, accuracy is of
the greatest concern, since it has a decisive influence on the quality of the subsequent digital imaging products. The
level of accuracy and reliability of direct georeferencing also influences the degree to which ground control points can
be eliminated.
This paper deals with the time synchronization issue for a direct georeferencing system based on integrated GPS/INS.
Time synchronization is generally considered to be one of the most critical factors in order to achieve high accuracy.
Fundamentally the issue is that the GPS and INS technologies were developed independently, hence their data refer to
separate internal clocks as time references. The necessity for higher time synchronization accuracy is directly related to
the increasing accuracy requirement for GPS/INS direct georeferencing systems addressing modern cutting-edge
applications.
This paper has analysed the time synchronization issues in GSP/INS integration. The impact of data synchronization
error and transmission latency on Kalman filtering results has been investigated using experimental data sets. Optimal
synchronization solutions have been proposed according to different integration scenarios. Although this study is based
on a GPS/INS integrated platform, the principles are equally applicable to the cases when more imaging and navigation
sensors are involved.
BIOGRAPHY OF PRESENTER
Weidong Ding is currently a Ph.D. student in the School of Surveying and Spatial Information Systems, the University
of New South Wales, Australia. He received his B.E. in Electrical Engineering from Beijing Polytechnic University,
China, and his M.E in Electrical Engineering from the University of New South Wales, Australia. His research is
focussed on developing an integrated positioning and georeferencing platform for kinematic positioning, to address
mobile mapping applications.
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
INTRODUCTION
More and more surveying companies are embracing digital technologies such as CCD cameras and LiDAR for mobile
aerial or ground based photogrammetric mapping and imaging applications. Advanced digital technologies have
provided a solution to the increasing demand for spatial information data in various applications ranging from surveying
and mapping to real-time location-based services. It has also reduced the overall cost of collecting, producing and
storing such data. Additionally, it brings advantages that conventional methods could not offer, such as 3D information
from LiDAR.
Direct georeferencing plays a crucial role in these developments. Direct georeferencing refers to the method of
providing the solution of the exterior orientation parameters of position, velocity and attitude on-the-fly without the use
of block adjustment procedures during post-processing [El-Sheimy, 2005]. The obvious benefit of direct georeferencing
is a reduction in the requirements for aerial triangulation and ground control points, both of which account for a
significant cost in survey work. Although methods like aerial triangulation have been advanced to automated aerial
triangulation during the past decade, the processing still needs a large amount of intensive editing by highly skilled
operators [Cramer et al., 2000].
In the case of direct georeferencing, accuracy is of the greatest concern, since it has a decisive influence on the quality
of the subsequent digital imaging products. The level of accuracy and reliability of direct georeferencing influences the
degree to which ground control points can be eliminated. Much research has been conducted in the surveying and
mapping field to investigate the influences of direct georeferencing accuracy on the quality of final imaging products
[Tachibana et al., 2004; Skaloud, 2002; Cramer et al., 2000; Kremer, 2002].
Direct georeferencing systems currently in use are mainly comprised of GPS and INS. Qualified delivery of exterior
orientation parameters depends on successful integration of GPS and INS subsystems. Time synchronization is
generally considered to be one of the critical factors in achieving high integration accuracy. Fundamentally the issue is
that the GPS and INS technologies were developed independently, hence their data refer to separate internal clocks as
time standards. This non-coordination results in the problem of time tagging bias, asynchronous sampling instants and
different sampling rates between GPS and INS.
Early studies of time synchronization issues include Bar-Itzhack’s [1984] observation of enigma bias during INS
transfer alignment. Knight [1996] described an accurate tagging of INS raw measurements with GPS time as probably
the single most critical element of successful GPS/INS tight coupling. Grejner-Brzezinska [2004] shows that time
synchronization is a factor that is crucial for achieving high accuracy positioning based on multi-sensor integration. Li
[2004] has successfully designed a cost-effective experimental device to mitigate INS data transmission uncertainty.
Due to its critical role in GPS/INS integration, a further comprehensive and systematic study of time synchronization is
needed to identify working mechanisms and to develop methods to mitigate time synchronization errors. This paper
attempts to provide some analysis on time synchronization error propagation rules and to provide useful information for
GPS/INS integration platform design.
In this paper, the factors influencing time synchronization accuracy are discussed. Then the tight coupling filter model
used in data processing is described, since specific discussion on time synchronization accuracy is closely related to the
filtering model used. The impact of data synchronization error and transmission latency on Kalman filtering results is
analysed using experimental data sets. Finally features and critical defining factors of reaching time synchronization
under different scenarios are discussed in detail, and some optimal solutions are proposed.
FACTORS INFLUENCING SYNCHRONIZATION
To demonstrate the time latencies in the GPS/INS integration context, a typical loosely coupled model is illustrated in
Figure 1.
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
GPS time t
Integration Platform
GPS receiver
t gps ? t ? e gpsbias
tins ? t free
INS
IMU
Sensors
Hardwiring 1PPS
GPS 1PPS capture time
RS232 Message
GPS Message capture time
Hardwiring 1PPS
INS 1PPS capture time
RS232 Message
INS Message capture time
t t?? eegpsbias ?? ee3
?e 7eof5 loosely
1
?gpsbias
Fig. 1 Clock bias and transmissionttins
latency
coupled mode
ins
A typical loosely coupled GPS/INS integration model is comprised of three functional subsystems, i.e. GPS receiver,
INS, and integration platform. All of the three subsystems need to have internal clocks for proper operations.
The GPS receiver, due to the nature of its design, has the possibility to link its internal clock time to GPS Time, which
is subsequently linked to UTC Time. Although there is a time difference between UTC Time and GPS Time, it is trivial
and not relevant to the issue of time synchronization accuracy in GPS/INS integration. So satellite time is considered as
the real time reference. Not every GPS receiver synchronizes its internal clock to satellite time all the time. It depends
??t ?t internal
t? ?t?gpsbias
e???TeT2gpsmess
on individual receiver design. Technically speaking, a GPS receiver can synchronize
its
to
time
tttins
e 5?7 ?e?1?e?e6eclock
T gpassatellite
gpsmess
gps1 pps
1 pps ? tins
1?pps
1 pps
insmess
insgpsbias
81 4? Tins
insmess
(GPS Time) to within 50ns accuracy.
In contrast, INS works rather independently, and has no link to real time references but is operating in a free-run mode.
This feature fundamentally creates the time synchronization problem for GPS/INS integration, because the data
measured by INS cannot match perfectly to GPS data by just using time tags as it would be possible in other
instrumentation systems. The only way the integration algorithm can do it probably is to align GPS and INS data
according to the instant when they are processed by the integration platform.
Transmission and processing latency add to time synchronization ambiguity. Since in most cases GPS and INS data are
transmitted using a serial communication link like RS-232(EIA-232), RS-422(EIA-422), time delay is unavoidable
especially when a buffering technique is used and/or the communication load is heavy. Data losses may happen due to
overflow when transmission load is heavy, data distortion caused by ambient interferences, and so on. Then the whole
correspondence relationship of the data is disrupted.
Furthermore, running an integration algorithm like a Kalman filter is computationally intensive. Data arriving at the
processing platform is usually put into a queue waiting for unpacking and error checking, before they are properly time
tagged by the processing platform using internal clock time.
The impact of the above time delay and uncertainty can be approximately illustrated using the loosely coupled model.
In a general sense, the common method of constructing integration algorithms is based on differencing GPS and INS
measurements to form error propagation rules from one epoch to the next. The navigation parameters and INS
modelling parameters are estimated using data fusion techniques like the Kalman filtering. In such a case, the dynamic
equation has the general form as follows:
eAewzHev=+=+&
(1)
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
Where
e
is the state vector of errors
A
is the dynamic matrix
H
is the observation matrix
()()()GPSINSztXtXt=−
z
is the observation vector which can be calculated by
,wv
are the noises that are assumed to be white and Gaussian. Their covariance matrices are Q and R respectively,
see Grewal et al. (1993) for more details about Q and R.
Time synchronization biases between GPS and INS data would introduce additional measuring errors:
()()()GPSINSztXtXtt=−+Δ
After making a Tailor expansion of
()INSXtt+Δ
, the observation equation can be expressed as
zHevζ=++
Where
ζ
(2)
(3)
is the additional residual caused by different data sampling time between the GPS and INS measurements.
ζ
When it is not explicitly addressed,
is often implicitly treated as a part of measuring noise in the modelling process.
ζ
However,
is not guaranteed to be white and Gaussian as required by the Kalman filter. In this case, additional
estimation error is introduced into the filtering results.
INTEGRATION FILTER
When Kalman filter is used for GPS/INS integration, it is often classified as either direct or indirect filtering [Hwang,
2004]. In direct filtering, total states such as position and velocity are chosen as state variables of the dynamic
equations, and direct measurements from GPS and the INS outputs are used as observables. Such integration filtering
architecture is mainly used in submarine applications and GPS stand-alone situations. The widely accepted integration
model in surveying for georeferencing is indirect filtering. Errors of position, velocity, acceleration and parameter error
of INS modelling are treated as state variables. The error states can be fed forward and back to compensate sensor errors
and to correct the navigation outputs.
In such a context, INS psi-angle error model or its equivalence is often used in the GPS/INS tight coupling model [BarItzhack, 1977; Bar-Itzhack et al.; 1988, Grejner-Brzezinska, 2004; Lee et al., 2001]:
()ieineninrä δωωδδψδδωδδψωδψε=−+×−×++∇=−×+=−×+vvfgrv&&&
(4)
δv,
∇ δr and δψ are the velocity, position, and attitude error vectors respectively
is the accelerometer error vector
δg
is the error in the computed gravity vector
ε
is the gyro drift vector
A twenty four state Kalman filter is used for data fusion and error estimation, which includes nine navigation solution
errors of position, velocity, and attitude in three dimensions, six accelerometer error modelling parameters (bias and
scale factors of each axes), three gyro drifts, three gravity uncertainty errors, and three lever arm values. Refer to
equation (5) for details.
[,,,,,,,,][,,,,,][,,,,,][,,][,,]TNavNEDNEDNEDTAccbxbybzfxfyfzTGyrobxbybzfxfyfzTGravNEDTAntbxbybzrrrvvvgggLLLδδδδδδδψδψδψεεεεεεδδδδδδ==∇∇∇∇∇∇
(5)
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
GPS measurements in the Kalman filter consist of double-differenced carrier phase observations:
CPH
GPS Base
CPH
GPS Rover
DD computation
Fixing ambiguity
DD error
KF
Predicted
DD
?V
? ?
IMU
Corrections
INS navigation
solution
Fig. 2 Tight coupling Kalman filter
DATA ANALYSIS
The AIMSTM software package was used for GPS/INS integration processing. It was developed by the Center for
Mapping at the Ohio State University (OSU) for large scale mapping and precise positioning applications (GrejnerBrzezinska, 2004; Da, 1997).
To investigate the impact of time difference on the positioning errors, two sets of ground based experimental data were
processed using AIMS. The first set of data was obtained from the Center for Mapping,OSU, while the second set was
collected at the University of New South Wales (UNSW). The hardware configuration for the system used by OSU for
collecting the data consisted of two GPS receivers, and one Litton LN-100 strapdown INS. The configuration used by
UNSW consisted of two Leica receivers and one DQI-NP INS. (DQI-NP is used purely as an INS, even though it has a
GPS receiver integrated with it.)
Processing results of the original OSU data are illustrated by Figure 3, which shows the 3D trajectory of the system.
e
e
r
g
e
d
n
i
225
220
t
h
g
215
i
e
210
H
-83.052
-83.047
-83.042
Longitude in degree
-83.038
39.997
39.998
39.999
40
40.001
40.002
40.003
40.004
Latitude in degree
Fig. 3 Output trajectory when OSU original data is processed
Since an accurate reference trajectory can not be directly measured, first the ambiguity-resolved segments are used for
comparison in order to assure the accuracy of reference at the centimetre level. Figure 4 shows the positioning error in
the local North-East-Down frame (NED) when compared with the GPS-only solution. The standard deviation of the
positioning errors is within 2cm. These results conform with the reported test accuracy from OSU [Grejner-Brzezinska,
2004; Da, 1997].
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
)
m
(
h
t
r
o
N
)
m
(
t
s
a
E
)
m
(
t
h
g
i
e
H
0.2
)
m
(
mean 0.017
std 0.015
h
t
r
o
N
0
-0.2
4.165
4.17
4.175
4.18
4.185
4.19
4.195
x 10
0.2
)
m
(
t
s
a
E
0
4.17
4.175
4.18
4.185
4.19
4.195
x 10
0.2
5
mean 0.005
std 0.017
4.17
4.175
4.18
4.185
4.19
)
m
(
t
h
g
i
e
H
0
-0.2
4.165
4.195
x 10
GPS second
mean 0.026
std 0.019
0
-0.2
4.165
4.17
4.175
4.18
4.185
4.19
4.195
5
mean -0.016
std 0.017
-0.2
4.165
0.2
x 10
0.2
5
mean -0.026
std 0.021
0
-0.2
4.165
4.17
4.175
4.18
4.185
4.19
4.195
x 10
0.2
5
mean -0.005
std 0.026
0
-0.2
4.165
4.17
4.175
4.18
4.185
GPS second
5
4.19
4.195
x 10
5
Fig. 4 Positioning error compared with GPS-only results (only the part
Fig. 5 Positioning error compared with GPS-only results when time
when ambiguities are resolved)
delay is 10ms
Then time latency was added to the time tags of the INS data in order to simulate transmission delay. The resulting INS
data were re-processed following the same integration procedure as before. As expected, an increase in positioning
errors has been observed in all the test results, with different time delays of INS data. To demonstrate, Figure 5 shows
the situation when a 10ms delay was simulated. It can be seen that positioning error increased, but the change did not
reach a significant level. 10ms delay only caused the standard deviation of positioning errors to increase to over 2cm.
The means of the positioning errors were also increased to more than 2cm.
As different increments of time delay were added to the time tags of the INS data, the positioning errors steadily
increased as the time delay became larger. The same trend was obvious even when the magnitude of the time delay
became negative, as for the case when INS data are transferred faster than GPS data (Figure 6). It can be concluded that
time tag differences of the order of 10ms between GPS and INS data seem tolerable when only positioning error is
considered, and when GPS phase ambiguities can be resolved.
s
a
i
b
0.35
r
.
o
ct
cc
A
a
0.3
r
e 0.25
t
e
m
n
i
f
e
l
a
c
s
0.2
.
c
c
A
D
T 0.15
S
s
a
i
b
0.1
0.05
0
-40
-20
0
20
40
60
80
Delay time in ms
Fig. 6 STD of positioning errors versus time delay
100
3
x 10
-3
2
1
0
-20
-10
0
10
20
30
40
50
-10
0
10
20
30
40
50
-10
0
10
20
30
40
50
0.2
0.1
0
-20
0.01
o 0.005
r
y
G
0
-20
delay time in ms
Fig. 7 Estimated INS sensor errors versus time delay
In order to investigate the impact of time delay on the trajectory segments when GPS phase ambiguities cannot be
resolved; INS errors estimated by the Kalman filter were further studied. In contrast to the segments when GPS phase
ambiguities were resolved and where positioning error can be somehow bounded by the high precision GPS
measurements, positioning accuracy for segments when the GPS phase ambiguities are unresolved is significantly more
degraded. The estimated accelerometer biases, scale factors and gyro biases when time delay was added to INS data
were compared with those when there was no time delay. Figure 7 illustrates the differences in the estimation of INS
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
sensor measuring errors from the no-delay estimation versus having delays. The estimation when there was no delay
was treated as reference, so difference is zero at delay time zero.
It can be seen from Figure 7 that increasing time delay has caused the estimated INS errors to deviate quickly from the
original estimation (when there is no time delay added). For
gµexample, the initial estimation of accelerometer biasgin
µ the
Kalman filter was 3.0e-4 m/s/s which equals to about 30
. The technical specification of the INS states 25
for
accelerometer bias. The 10ms delay has caused additional estimation error more than that amount. This degradation of
estimation accuracy may explain the large error when the complete trajectory with 10ms time delay as shown in Figure
8 is compared with the original trajectory in Figure 3.
r
o
r
r
e
Integrated resolution
e
d
u
t
ri
t
o
ra
rL
e
e
e
r
g
e
d
225
n
i
220
t
h
g 215
i
e
210
H
-83.052
-83.05
-83.048
40.004
-83.046
40.002
-83.044
40
-83.042
39.998
-83.04
Longitude in degree
39.996
e
d
u
t
i
g
n
r o
oL
r
r
e
t
h
g
i
e
H
4
x 10
-6
2
0
-2
4.165
4.17
4.175
4.18
4.19
x 10
4
x 10
5
2
0
-2
4.165
4.17
4.175
4.18
4.185
4.19
x 10
5
0.2
0
-0.2
4.165
4.17
4.175
4.18
Latitude in degree
Fig. 8 Output trajectory when 10ms delay is added to INS data
4.185
-6
4.185
4.19
x 10
5
Fig. 9 Error of trajectory with 10ms time delay when compared to
original trajectory
10ms delay in the INS data has caused the integration accuracy to degraded from the centimetre level to the decimetre
level when GPS phase ambiguities are not resolved in a short period of time, i.e. max. 111 seconds (Figure 9). When
GPS phase ambiguity resolution is lost for a longer time, for example the last part of the trajectory has lost ambiguity
resolution for about 20 minutes, the resulting trajectory diverged from the referenced one very quickly. (The divergent
part is not drawn out in Figure 9 in order to show more details of the smaller errors.) It should be noted that the
magnitude of the errors demonstrated here is only in a relative sense since the comparison is not based on the true
trajectory.
Correlation coefficients between velocity trajectories and positioning error trajectories were calculated at different delay
times in order to evaluate the influence of vehicle speed on GPS/INS positioning accuracy. No obvious correlation was
detected at any of the tested delay steps (Figure 10).
SYNCHRONIZATION METHODS
To solve the time synchronization problem systematically, three scenarios have been identified [Li, 2004] according to
the nature of the GPS/INS configurations. Then solutions are proposed accordingly.
In the following discussion, the GPS receiver is always considered as the time master which provides a 1PPS timing
signal and serial timing messages. It is also assumed that the GPS-derived 1PPS indicates the physical time of the
validation of the GPS timing messages.
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
s
t
n
e
i
c
i
f
f
e
o
c
0.1
Temprature
Sensor
0.09
0.08
Gyros
Multiplexer
0.07
n
o
i
t
a
l
e
r
r
o
C
A/D
Converter
Data output
Sample/
Hold
0.06
Accelerotmeters
0.05
0.04
0.03
Smapling control circuit
GPS 1PPS signal
0.02
0.01
0
-40
-20
0
20
40
Delay time in ms
60
80
100
Fig. 11 Block diagram of IMU data sampling circuits [Ma et al., 2004]
Fig. 10 Correlation coefficients between velocity and positioning
error
versus time delay
Scenario one - synchronization at INS’s data sampling circuit
In terms of accuracy, the best solution is to implement time synchronization in the data sampling circuits. It is also the
common method in the instrumentation industry when high precision of synchronization of data sampling is required. A
common reference sampling clock (or counter) is used to synchronize the sampling action, normally implemented in the
A/D circuits.
The sampling control circuit creates triggers according to sampling frequency required. Each trigger is used to start
sample and hold circuits to take a snapshot of all the sensor measuring values at exactly the same time. Meantime an
internal counter value is frozen as an indication of the sampling time. In the A/D output, the digital value is combined
with the counter value to form a complete measurement with precise corresponding time. GPS 1PPS signal is
introduced to sampling control circuits to control the generation of sampling pulse.
Although high accuracy can be achieved using this method it is not an easy task to modify INS hardware circuits or to
develop an INS from individual IMU sensors.
Scenario two - synchronization using GPS and INS’s timing signals
The characteristic of this scenario is that the INS works like a “black box”, but with pulses output indicating the
physical time of validation of timing messages.
1e 5eThe relationship between GPS 3and
e 7eINS clocks can then be traced. This
is
the
case
demonstrated
in
Figure
1,
where
,
are
1PPS
alignment
errors,
.
, are timing message output delays,
2e 4e 6e 8e
,
,
,
are processing delays within the integration platform.
Since the magnitudes of
e1 and e5 are available from product technical specifications, and e2 and e6 should have the
same magnitude depending on integration platform performance, we have
11()insinsppsgpsppsttTT≈+−
(6)
According to the correspondence between the INS timing pulse and time message, the relation between INS clock time
and GPS Time can be found. Time tagging of INS data can then be translated into GPS Time.
The above equation shows that the clock error and drift of the integration platform does not influence
tins much, since
its error is not cummulative. When the short term stability and accuracy of the clock oscillator is satisfactory, the GPS
message and INS data can be well synchronized by checking the time difference between the GPS 1PPS time and the
INS pulse time. A sufficiently accurate and stable digital counter can be used to calculate the time difference between
the GPS 1PPS and INS pulses.
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
Since sampling instants are not coordinated between GPS and INS, guaranteed accuracy depends on the INS sampling
rate. When INS has a sampling rate of 100Hz, guaranteed synchronization accuracy is 0.005 seconds. This is one
drawback when compared with method introduced in scenario one. Sometimes an interpolation method can be used to
improve the “virtual” sampling rate of INS.
Some GPS receivers provide a function called event time tagging. It is not suitable for synchronizing INS data because
it is hard to match the INS data and the recorded pulses’ time tags during processing when the sampling rate is high.
Scenario three - synchronization without INS’s timing signals
Many INS or IMU products do not provide the timing pulses described in scenario two. In this case time
synchronization has to be performed by analysing the digital signal features.
Often IMU/INS units are using serial communication to output measuring data, most likely RS232 (EIA232), RS422
(EIA422), 1553B, etc. In these cases, factors influencing synchronization accuracy include sampling and processing
delay in the IMU/INS, serial communication delay, buffering uncertainty, and time differences of the sampling instant.
Measures can be taken to reduce transmission delay and uncertainty. The best is to find a time mark which can
definitely indicate the beginning of the transmission, so the transmission time can be estimated and compensated for.
For example, hardware handshaking signals for serial communication can be used for this purpose. Measures to prevent
buffer overflow at the serial port include having a link speed high enough, and giving the service program higher
priority. INS communication load is easy to estimate.
Manufacturers may provide parameters concerning the internal sampling and processing delays of their INS. Otherwise
the value has to be determined in a laboratory. The correlation method can be used to check the correspondence of
trajectory features of the INS navigation output, and the GPS positioning measurements. The techniques are similar to
those used in airborne vector gravimetry [Kennedy, 2002; Kwon, 2000].
Nevertheless, higher time synchronization accuracy cannot be expected in this case when compared with scenarios one
and two.
CONCLUSION
This paper has reviewed the time synchronization issue in GPS and INS integration in the context of direct
georeferencing applications. Factors influencing the accuracy have been identified and analysed. The impact of time
synchronization error on GSP/INS positioning accuracy has been studied with the aid of experimental data sets.
The test results have demonstrated that within the integration Kalman filter, the measuring errors caused by the time
synchronization biases are not directly transferred into positioning errors. However, millimetre level synchronization
accuracy is still necessary in order to calibrate the INS parameters properly. Three time synchronisation scenarios have
been described. The feasibility of each of these scenarios has been discussed .
It has been noted that high dynamics of the platform does not necessarily increase the impact of the time
synchronization biases on positioning errors. The impact of time synchronization errors on positioning accuracy
depends on the GPS/INS integration model and the Kalman filter structure. This needs to be further investigated.
REFERENCES
Bar-Itzhack, I.Y. (1977) “The Psi-angle error equation in strapdown inertial navigation systems”, IEEE
Trans.Aerospace and Electronic Systems, vol AES-13, 6, 679-689.
Bar-Itzhack, I.Y., Vitek, Y. (1984) “The enigma of false bias detection in a strapdown system during transfer
alignment”,. J.Guidance, vol 8, 2, 175.
Bar-Itzhack, I.Y., Berman, N. (1988) “Control theoretic approach to inertial navigation system”, AIAA Journal of
Guidance, Control & Dynamics, vol 11, 3, 237-245.
Time Synchronization Design for Integrated Positioning and Georeferencing Systems
El-Sheimy, N. (2005) “An overview of Mobile Mapping Systems”, Pharaohs to Geoinformatics FIG Working Week
2005 and GSDI-8, Cairo, April 16-21.
Cramer, M., Stallman, D., Haala, N. (2000) “Direct georeferencing using GPS/Inertial exterior orientations for
photogrammetric applications”, IAPRS, vol XXXIII, Amsterdam.
Da, R. (1997) “Analysis and test results of AIMS GPS/INS system”, proceedings of ION GPS-97, Kansas City, USA,
16-19 September, 771-780.
Grejner-Brzezinska, D.A. (2004) “Workshop Notes of Integrated System for Mobile Mapping”, workshop presented
at the University of New South Wales, 13-14 December.
Grewal, M.S., Andrews A.P. (1993), Kalman filtering theory and practice, Pretice-Hall inc., US.
Hwang, D.H. (2004) “Kalman filtering – fundamentals and applications”, seminar presented at the University of New
South Wales, 15th September.
Kennedy, S. (2002) “Precise acceleration determination from carrier phase measurements”, ION GPS 2002, Oregon,
USA, 24-27 September, 962-972.
Knight, D. (1996) User Manual of GINI Version 2.0, Knight Systems, US.
Kremer, J. (2002) “CCNS/AEROcontrol – an Integrated GPS/IMU system for direct georeferencing of airborne image
data”, symposium Gyro technology 2002, H.Sorg(Ed.), University of Stuttgart, Germany, 17-18 September, 16.016.9.
Kwon, J.H. (2000), Airborne vector gravimetry using GPS/INS, PhD thesis, Report No. 453, Dept. of Civil and
Environmental Engineering and Geodetic Science, the Ohio State University, USA
Lee, H.K., Wang, J., Rizos, C., Grejner-Brzezinska, D., Toth, C. (2002) “GPS/pseudolite/INS: Concept and first tests”,
GPS Solutions, 6(1-2), 34-46.
Li, B. (2004) “A cost effective synchronization system for multisensor integration”, ION GNSS 2004, Long Beach,
USA, 21-24 September, 1627-1635.
Ma, Y.F., Zhou, B.L., Wan Z.G.., Zhao, L.Y. (2004) “Data synchronization and fusion method in MIMU/GPS
integration navigation system”, Journal of Chinese inertial technology, vol 12, 3, 28-31.
Skaloud, D. (2002) “Direct georeferencing in aerial photogrammetric mapping”, photogrammetric engineering &
remote sensing, March 2002, 207.
Tachibana, K., Sasagawa, T., Yotsumata, T. (2004) “Experimental study of direct georeferencing system (DGS) for
large-scale mapping”, the XXth ISPRS congress, Istanbul, 12-23 July, 860 ff..
View publication stats