Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Estimation of Multipath Range Error for

Detection of Erroneous Satellites


Tatsuya Iwase, Toyota Central R&D Labs., Japan
Noriyoshi Suzuki, Toyota Central R&D Labs., Japan
Yusuke Watanabe, Denso Corp., Japan

One of the previous technologies for tackling this problem


BIOGRAPHIES is RAIM, which checks the residual of the least square
and identifies the suspicious satellites. However, it
Tatsuya Iwase is a Researcher in the ITS laboratory of presumes a Gaussian measurement error which is more
Toyota Central R&D Labs. He received his B.E. and M.E. common in an open-sky environment than in the urban
degrees from the University of Tokyo in 1998 and 2000, canyon environment.
respectively. His current research focuses on precise On the other hand, our proposed method tries to estimate
GNSS positioning of land vehicles in urban environments the size of the multipath error directly and use the
and its safety applications. estimated error to weight satellite signals and improve the
positioning accuracy.
Noriyoshi Suzuki is a Senior Researcher in the ITS The proposed method was tested with a low-cost GNSS
laboratory of Toyota Central R&D Labs. He received his receiver mounted on a test vehicle in a test drive in
B.E. and M.E. degrees in electronics engineering from Nagoya, Japan, which is a typical urban canyon
Kyoto University in 1990 and 1992, respectively, and the environment. With this experimental data, the accuracy of
D.E. degree from Toyohashi University of Technology in the multipath error estimation and of the improved
2003. His research interests include digital mobile positioning are discussed.
communication systems and integrated GPS/INS systems
for land vehicle positioning. 1. INTRODUCTION

Yusuke Watanabe is a Researcher in the ITS For land vehicle navigation, several positioning sensors
Development Dept of Denso Corp. He received his B.E. are used such as GNSS and INS, and their measurements
degrees from Nagoya Institute of Technology in 2000. are integrated to achieve higher accuracy of positioning
His current research focuses on precise combined by compensating for the shortcomings of each
GNSS/INS positioning of land vehicles. [1][2][3][4][5][6]. However, the blockage and the
reflection of satellite signals in urban canyons cause
multipath pseudorange errors and degradation of the
ABSTRACT GNSS positioning performance. Especially, if the direct
signal is blocked by buildings and only indirect signals
In integrated systems for accurate positioning which reach the GNSS receiver, the pseudorange error becomes
consist of GNSS, INS and other sensors, the GNSS so large that GNSS lowers the positioning accuracy of the
positioning accuracy has a decisive influence on the entire system. Thus, it is important to detect false
performance of the entire system and thus is very multipath satellite signals that cause pseudorange errors
important. However, GNSS usually exhibits poor and exclude them from positioning calculations to
positioning results in urban canyon environments due to maintain positioning accuracy.
pseudorange measurement errors caused by multipath One of the previous technologies tackling this problem is
creation, which leads to performance degradation of the receiver autonomous integrity monitoring (RAIM)
entire positioning system. For this reason, in order to [7][8][9][10][11]. RAIM detects signals from GNSS
maintain the accuracy of an integrated positioning system, satellites which are erroneous and subsequently cause
it is necessary to determine when the GNSS positioning is mistaken positioning by statistical reliability tests. RAIM
accurate and which satellites can have their pseudorange is a type of robust estimation, which is a technique to
measured accurately without multipath errors. detect errors, namely outliers, based on calculation of the
Thus, the objective of our work is to detect the multipath least square residuals. In epoch-by-epoch GNSS
errors in the satellite signals and exclude these signals to positioning, the least square method is commonly used, in
improve the positioning accuracy of GNSS and which the receiver position which best maintains
consequently the integrated positioning system, especially consistency with pseudorange measurements of satellites
in an urban canyon environment. is estimated. The least square residual of a satellite is the
discrepancy between this estimated position and observed

2228
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
pseudoranges of the satellite, and reveals inconsistency of specifically pseudoranges, from the errors in solutions,
this satellite’s signals with others. RAIM regards an specifically vehicle positions.
observation with a large residual as an outlier, and The solution equations of GNSS positioning have four
excludes it from the positioning calculations. variables: vehicle position (x,y,z) and receiver clock bias.
However, if the percentage of outliers among observed This requires four references to compute the errors in the
signals increases, the estimated receiver position solution. However, if the four reference values are
gravitates toward those outliers and away from the true available, that means the positioning problem is already
position. This results in the reduction of the residuals of solved. Usually only some of the four references are
outliers and increase in the residuals of proper available. This insufficiency of references causes this to
observations, and the subsequent failure of the reliability be an underdetermined problem. In this paper, a technique
tests. The threshold percentage of outliers causing the to solve such problems is proposed.
failure of robust estimation is called breakdown point. First, an equation is derived which shows the relation
Generally, the breakdown point is up to 50%. Therefore, between the error in the solution and the error in the
if the percentage of erroneous satellite signals exceeds observation. The GNSS solution is calculated with the
50%, RAIM will fail. Such a situation is likely in urban Newton-Raphson method as follows
environments where high buildings are concentrated and ∆x = x − x 0
signals from satellites are often blocked. In these severe
environments for GNSS, an alternative technology to ∆ri = ri − x 0 − x i
traditional RAIM is necessary to correctly choose from
among observed satellite signals. ∆r = H∆x
In this paper, we study a new approach by which where x is the solution to be estimated, x 0 is the initial
erroneous satellite signals are identified and excluded to
value of the solution, x i is the position of the satellite
improve positioning accuracy in urban environments.
Unlike RAIM which calculates least square residuals, our number i, ri is the pseudorange of the observed satellite
proposed method directly estimates the size of the signal with ionospheric delay, tropospheric delay and
multipath error in the observed signal based on partially
satellite clock bias eliminated, ∆r is the vector
available reference data of vehicle position.
The rest of this paper is as follows. Section 2 describes representation of ∆ri , and H is a geometry matrix. ∆x
the basic concept of our method for estimating the is calculated with the least square method as follows.
multipath errors in observed signals. Section 3 and 4 ∆x = H + ∆r
describe error estimation with an altitude map. Section 5
describes an experiment to evaluate the performance of H + is a pseudo-inverse matrix of H calculated as
the proposed technique and its results. Finally, follows.
conclusions are presented in Section 6. H+ = HTH( )−1
HT
If the pseudorange includes multipath error e , it causes
2. BASIC CONCEPT OF MULTIPATH
ESTIMATION the solution error e x as follows.
∆x + e x = H + (∆r + e)
As described above, RAIM fails in GNSS reliability tests
when more than half of the observed signals are incorrect. ∴e x = H +e
This is because RAIM tests the observations by The elements of this equation are as follows.
calculating least square residuals. Here, we propose a new
approach to reliability testing which uses reference  ex   a1 an 
information of vehicle position instead of least square e   b   e1 
 M 
b
residuals. If the size of observed signal error is known, it  =
y 1
L n

is easy to discriminate the erroneous satellite signals. The  ez   c1 cn   


    e 
size of errors is calculated as the difference between
 eb  d1 dn  n 
observations and true values. Usually the true value is not
available, and the most reliable and precise value is used In this paper, we call this the error equation. With this
as the reference instead. equation, the observed signal error e can be calculated
In GNSS positioning, the solution (the position of the from the solution error e x . Note that the error equation
vehicle) is estimated from the pseudorange observations.
does not make any assumption regarding the error, such
If a reference pseudorange is available, the size of errors
as Gaussian distribution. Accordingly, the error equation
in pseudorange observations can be calculated and the
is useful in analyzing non-Gaussian noise such as
erroneous satellite signals can be detected easily.
multipath errors.
Here is an example of error calculation. Let ( x, y, z , b )
However, it is difficult to obtain such a reference
pseudorange. On the other hand, reference vehicle
positions are sometimes available, and at those times the denote the GNSS solution of the vehicle position and
error in vehicle position can be estimated. The basic idea receiver clock bias. Now let's assume that only the vehicle
of our approach is calculating the errors in observations,

2229
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
position has a reference value, which is ( xr , yr , zr ) . reliable and its signal error is zero. We call this satellite
the reference satellite. This is denoted as follows.
Then the error in positioning is as follows.
 e x   x − xr   e1 
    M
 e y  =  y − yr   
e   z − z  0 = [0 L 1 L 0] ei 
 z  r  M
According to the error equation, the relation between the e 
error in positioning and the error in pseudorange  n
observation is as follows. With this equation added in the simultaneous equation
above, the following new equations are set up.
 ex  a1 an   e1 
e  =  b L b   M   ex   a1 an −1 an 
 y  1 n   e   b L b bn 
 ez   c1 cn  en   y  1 n −1

cn   1 
 ez   c1 e
The element of clock bias is omitted because it does not
cn −1
M  M  M
have a reference. This equation applies to the case where
e n  = a n  
satellites 1 to n are used in positioning. If the number of n an −1 0 en −1 
 xn   1n 
observations is n, which is larger than 4, the solution of
e y  b1 L n bn −1 0  en 
the position can be calculated with different combinations
 n  n 
of the satellites. If we use all the satellites and remove one
ez   c1 n cn −1 0
of the satellites one by one, there are n+1 combinations.
For each combination, the solution can be calculated and
 0   0 L 1 L 0 
As a result, the rank of the right side matrix becomes n,
the error equation can be set up. For example, the error
equation without the first satellite is as follows. making the simultaneous equations solvable. In vector
representation, the equations are expressed as follows.
e  e xyz = B xyz e
 1 ex  0 1 a2 1 an   1 
 e  = 0 L  e2  Using the least square method, the pseudorange error can
1 y   1 b2 1 bn 
M be calculated as follows.
 1 ez  0 1 c2
 
1 cn  
+
e = B xyz e xyz
en 
If the speed of the vehicle has a reference value, the same
Note that the elements in the matrix of the error equation
technique can be adopted to estimate the Doppler
are different for each satellite combination. The suffixes
observation error. As one example, Fig. 1 shows the
at the lower left mean the satellite number which is
pseudorange errors estimated during an experiment in an
excluded in the positioning calculation.
urban environment in Japan. The details of the experiment
With all the n+1 error equations of different satellite
are explained in a later section. Fig. 2 shows the sky plot
combinations combined, the following simultaneous
of satellites at this time. It is clear that the size of
equations are set up.
pseudorange errors becomes large when a satellite is
 ex   a1 an − 1 an  invisible, blocked by buildings.
e  b L b bn   e 
During the error estimation process described above, we
 y   1 n −1
1
made the assumption that the reference satellite has no
 ez   c1 cn −1 cn   M  pseudorange error. If the reference satellite is blocked by
 M = M    eq. 1 a building and its observation includes multipath error,
 n ex   n a1 0 en −1 
the assumption is not true and the estimation fails. To
n an − 1
     prevent this failure, a simple test for multipath error can
 n e y   n b1 L n bn −1 0  en  be conducted by checking the stability of the receiver
 e  c 0
clock bias. By subtracting the estimated error from the
n z n 1 n cn −1  observed pseudorange, a precise observation and the
positioning solution can be obtained without the influence
of multipath. The receiver clock bias in the solution
The observed signal errors can be obtained by solving
should be also out of the multipath influence and then its
these simultaneous equations. However, the rank of the
time series should be smooth. If the reference satellite
matrix on the right side is n-1, and the simultaneous
signal includes multipath error, it distorts the clock bias
equations need one more constraint to be solved. This is
solution and the failure of error estimation can be detected
because the simultaneous equations only have the
as this disturbance of the clock bias. Fig. 3 shows the time
constraints of the spatial position, and information on
series of the second derived clock bias solution. The
clock bias is missing. For this reason, another assumption
disturbance of multipath appears as large jumps in the
needs to be introduced as a constraint. Here, we assume
graph. During this time period, a highway bridge was
that the satellite i with the highest elevation angle is
above the test vehicle and the reference satellite was

2230
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
blocked. Except for this period, it can be said that the In the last section, an error estimation technique is
estimation was correct. proposed using reference information of vehicle position.
However, finding the position of the vehicle is generally
the purpose of the GNSS positioning, and reference
Pseudorange Error [m]

positions are usually unavailable. Meanwhile, the altitude


of land vehicles basically depends on the shape of the
terrain. Reference altitudes are thus relatively easy to
obtain, especially where the slope of the road is not so
steep. Therefore, we used an altitude map as a reference
giving the vehicle altitude. This map is expressed as
values of the following function which takes longitude
and latitude values as inputs and outputs the altitude at
that point.

Satellite ID
z r = f ( x, y )
In a relatively flat area as found in most big cities, the
Fig. 1 Estimated pseudorange errors error in the output altitude should be relatively small
compared to the error in the input horizontal position.
Accordingly, this altitude map can be an effective
reference in urban area positioning.
By using this altitude map as a reference, the same basic
procedure described in the previous section can be used to
estimate the error of GNSS observations. Since only
altitude reference values are available here, Eq.1 is
changed into equations only dealing with altitude as
follows.
 ez   c1 cn 
 e1   0   e1 
 z =  1 cn   
L M
 M   n −1 c1 n −1 cn
 
 n   en 
ez   n c1 0 
The rank of the right side matrix is n-3 this time, since
this equation has only altitude information as a constraint.
Thus, three reference satellites are necessary to solve the
equation. Once the simultaneous equations are set up with
values from the three reference satellites, the rank of its
matrix becomes full and the equations are solvable as
follows.
Fig. 2 Skyplot of the error estimating situation e = Bz+e z
This estimated error of an observed signal is reliable if
clock bias [m/s2]

the assumption that the observed reference satellite


signals are not erroneous is correct. The truth of this
Second derived

assumption must be verified before we choose satellites


whose signals are used for positioning, based on the
estimated errors of their observed signals.
Fig. 4 gives an overview of the verification process flow.
First, in the assumption generation process, the most
reliable three satellites are provisionally chosen and it is
assumed that signals received from them do not contain
errors. Since satellites at high elevation angles are
Time [s] unlikely be blocked by buildings, they are more reliable
than satellites at lower elevation angles. Thus initially, the
Fig. 3 Time series of clock bias solution four satellites with the highest elevation angles are chosen
and positioning is calculated with signals from those
satellites. The altitude error in the solution is calculated
3. MULTIPATH ESTIMATION WITH with the altitude map as reference. If the altitude error is
INCOMPLETE REFERENCES less than a threshold value, those four satellites are
regarded as reliable, and the three with the highest
elevation angles are chosen as the reference satellites.

2231
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
Then, in the estimation process, the error of an observed
signal is estimated using the altitude map and the three 4. ASSUMPTION-VERIFICATION LOOP
reference satellites.
In the corrected positioning process, the estimated error is In the assumption generation process, the reference
then used to make a correct positioning calculation. This satellites are chosen based on their elevation angles.
can be done in several ways. The simplest way is However, when vehicle goes under a highway or is
deducting the estimated error from the observation to surrounded by high buildings, even satellites with high
cancel the influence of multipath. However, this yields the elevation angle are not always reliable. In such a situation,
same result as positioning with altitude map and three another technique is needed to choose appropriate
reference satellites. What we want is an optimum solution reference satellites. On the other hand, the proposed
with all the satellite signals taken into account except for technique described in the last section calculates an error-
the ones with multipath errors. The second simplest way corrected solution and verifies it with the altitude map.
to do this is to select satellites based on the size of the This verified positioning solution is reliable and can be
estimated errors. Another way is to weight satellites in used as reference information in place of high elevation
positioning according to the estimated error of their satellites. Fig. 5 shows a process flow using a verified
signals. This is done by a weighted least square technique solution as a reference. In the prediction process, a
[12]. For example, the weight is defined as follows. positioning reference of the current time is calculated

(e ≥ 1)
from a verified solution of a past time as follows,
 1 / ei x1* = x 0 + v 0 dt
wii =  i

1 − log( ei ) (e i < 1)
eq. 2
b1* = b0 + d 0
where x 0 , v 0 , b0 , d 0 are the past verified position,
where ei is the estimated error of satellite i.
*
The positioning solution is calculated with the following speed, clock bias and clock drift, respectively, and x1 ,
equation b1* are the predicted reference position and clock bias,
(
∆x = H T WH H T W∆r)−1
respectively. The speed in the past time is obtained either
by estimating Doppler error or by correcting the speed
where W is a weight matrix, the element (i,i) of which
solution by pseudorange error weighting. With these
denotes wii . predicted references, the error equation is set up. Now
Finally, in the verification process, the correctness of the that there are four references, full rank simultaneous
reference satellite signals is tested by checking the equations can be set up, and then no reference satellites
altitude error in the solution with the altitude map. If the are necessary. The observed signal error is estimated by
altitude error is less than a threshold, the assumption that solving these simultaneous equations. Since the
the signals of the three reference satellites are correct is references are calculated from the past verified
upheld. Otherwise, the assumption is rejected and the information, this estimated error is also based on the past
positioning does not output any solution since there are information. Thus, this estimated error is used only for the
not enough reliable satellites. This causes the reduction in selection of reference satellites at the current time in the
availability of positioning solutions throughout the urban assumption generation process. The three satellites with
area. the least estimated error are chosen as reference satellites,
and then the error estimation is conducted again with the
Assumption Generation:
Three reference satellites
selected reference satellites and the altitude map. Since
are chosen. the reference satellites supply current information, they
are more reliable than the predicted references based on
the past information. By using the past information, three
Estimation: reference satellites can be selected properly even when
Observation error is high elevation angle satellites are blocked.
estimated.

Corrected Positioning:
Positioning solution is
calculated with error
taken into account.

Verification:
Solution is tested
with altitude map.

Fig. 4 Flow of verification

2232
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
with and without correction using the estimated error
Assumption Generation:
Three reference satellites
information. In the positioning without correction, an
are chosen. elevation mask of 10 degrees and S/N mask of 25dB were
applied. In the positioning with correction, the estimated
error is used to weight satellites as in Eq.2. Compared to
Estimation: the positioning without correction, the corrected
Observation error is positioning exhibited improved positioning accuracy. The
estimated. Prediction: availability of positioning solutions in the urban area was
Reference positions decreased after the corrected positioning since some of
are calculated. the corrected positions were rejected in the verification
Corrected Positioning: process. As seen in the pictures, the positioning solutions
Positioning solution is with large error were correctly excluded.
calculated with error Another experiment was conducted with Gaussian noise
taken into account.
( )
N 0,10 2 added to the altitude map to see the influence
of the accuracy of the altitude map. Fig. 10 shows the
Verification: results. Results were not much different from Fig. 7, and
Solution is tested RMSE was 27.6m. These results indicate that the
with altitude map. proposed technique has robustness against the noise in an
altitude map. One of the possible reasons is that the noise
Fig. 5 Flow of assumption-verification loop
in altitude map is relatively small compared to the
pseudorange error, small enough to be ignored. The order
of the pseudorange error in an urban area is over 100m as
5. EXPERIMENT AND RESULTS shown in Fig. 7.
The positioning accuracy shown in Fig. 9 can be
An experiment was conducted with a test vehicle to improved if severe threshold parameters of multipath
evaluate the performance of the proposed error estimation detection are applied. In such case, the availability of
technique described in the previous section. The positioning solutions decreases in return for increase in
experiment was conducted on September 2, 2009 in the accuracy. Fig. 11 shows a result when the threshold of
Nagoya area in Japan. (The approximate location is pseudorange error is changed. The more stringent the
35.17N, 136.88E.) It is a dense urban area with buildings multipath detection standards get, the greater the accuracy
over one hundred meters high. The test vehicle used in and the lower the availability of positioning solutions. For
this experiment was equipped with a u-blox Antaris 4 comparison, the results of traditional RAIM are shown in
GPS receiver and POS LV, a high-end, precise the same graph. It is clear that the proposed technique can
positioning system.[13] make the accuracy of positioning solutions better than
Fig. 6 shows the process flow of the experiment. In the RAIM while keeping their availability as high as RAIM.
beginning, data of POS LV is used as the reference With more severe parameters, accuracy up to 11m can be
vehicle positions ( xr , yr , zr ) . With this reference data, attained even in urban areas.
error in GPS observation is estimated using the technique
proposed in Section 2. This estimated error is used as a GPS Observation
reference to verify the accuracy of error estimation using (Pseudorange,Doppler)
the altitude map, as described in section 3 and 4. This Altitude Map POSLV Data
time, the altitude zr measured by POS LV is used for the
output of the altitude map. As shown in Fig. 3, the
Error Estimation Error Estimation
reference error is not calculated correctly in some places. as in section 3&4 as in section 2
In those places, the performance evaluation was not
conducted due to the lack of reference data. The
parameters in error estimations were as follows. Estimated Reference
Comparison1
Error Error
• Threshold of altitude error in verification process:
40 meters
• Threshold of pseudorange error in assumption
Corrected Positioning
generation process: 5 meters Comparison2
Positioning Solution
• Sampling frequency in prediction process: 1
second Fig. 6 Flow of experiment
The result of the error estimation is shown in Fig. 7. The
x-axis shows the error estimated with altitude map, while
the y-axis shows the reference error. The estimated error
matched the reference error well. The RMSE was 27.5m.
Then, the estimated error was used to correct positioning.
Fig. 8 and Fig. 9 show comparisons between positioning

2233
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
Reference Error [m]
Reference Error [m]

Estimated Error [m] Estimated Error [m]


Fig. 7 Estimated pseudorange error Fig. 10 Estimated pseudorange error with noise in altitude
map

Reference Position
Proposed Method
GPS Solution 120 RAIM

2drms Error[m]
100
80
60
40
20
11m
0
0 20 40 60 80 100
Availability[%]
2drms Error:105.3m
Availability:100% Fig. 11 Relation between accuracy and availability

Fig. 8 Positioning error and availability


6. CONCLUSIONS

Reference Position A new approach to identify erroneous satellite signals and


exclude them for precise GNSS positioning was described.
GPS Solution The technique uses an altitude map as partial reference
information to test the reliability of satellite signals and
estimate the errors in the observed satellite signals,
including those due to multipath. The proposed method
was examined in an urban canyon where large multipath
error occurs. By this method, the size of pseudorange
error in observed signals was estimated for each satellite.
The estimation of the error was accurate enough that the
multipath satellite signals could be excluded correctly,
and the accuracy of the consequent positioning was
improved. It was also shown that the positioning accuracy
2drms Error:31.7m based on the proposed method was better than that based
Availability:75.1% on traditional RAIM with the same positioning solution
availability. The remaining problem of this technique is
that it assumes the reliability of the signals of three high
Fig. 9 Positioning error and availability (corrected) elevation angle satellites at the beginning of its algorithm.
The algorithm thus is faulty if the high elevation satellites
are blocked as when the vehicle is under a highway
overpass. This problem may be solved if S/N information
is used in a future system.

2234
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010
[13] http://www.applanix.com/products/land/pos-lv.html
ACKNOWLEDGMENTS

This study has been partially funded by Denso


Corporation. The authors are grateful for the advice of
project team members: Mr. Katsuhiko Mutoh, Mr.
Tadashi Kamada, Mr. Hiroyuki Kitagawa, and Mr. Kei
Uemura from Denso Corporation, and Ms. Yoshiko
Kojima, Mr. Junichi Meguro, and Mr. Kojiro Takeyama
from Toyota Central R&D Labs., Inc.

REFERENCES

[1] I Skog, P Handel, In-Car Positioning and Navigation


Technologies - A Survey, IEEE Transactions on
Intelligent Transportation Systems, Vol.10, No.1,
pp.4-21, 2009
[2] I Skog, P Handel, A low-cost GPS aided inertial
navigation system for vehicle applications, Proc.
EUSIPCO 2005, 2005
[3] F Caron, M Davy, E Duflos, P Vanheeghe, Particle
Filtering for Multisensor Data Fusion With
Switching Observation Models: Application to Land
Vehicle Positioning, IEEE Transactions on Signal
Processing, Vol.55, No.6, pp.2703-2719, 2007
[4] I Miller, B Schimpf, M Campbell, J Leyssens,
Tightly-coupled GPS / INS system design for
autonomous urban navigation, IEEE/ION Position,
Location and Navigation Symposium, pp.1297-1310,
2008
[5] A Soloviev, Tight coupling of GPS, laser scanner,
and inertial measurements for navigation in urban
environments, IEEE/ION Position, Location and
Navigation Symposium, pp.511-525, 2008
[6] M Jabbour, P Bonnifait, Backing up GPS in urban
areas using a scanning laser, IEEE/ION Position,
Location and Navigation Symposium, pp.505-510,
2008
[7] H Kuusniemi, G Lachapelle, JH Takala, Position and
velocity reliability testing in degraded GPS signal
environments, GPS Solutions, Vol.8, No.4, pp.226-
237, 2004
[8] CL Young, GO Daniel, A Performance Analysis of a
Tightly Coupled GPS/Inertial System for Two
Integrity Monitoring Methods, Navigation, 2000
[9] MG Petovello, ME Cannon, G Lachapelle, Kalman
filter reliability analysis using different update
strategies, Proceedings of the CASI Annual General
Meeting, 2003
[10] H Leppakoski, H Kuusniemi, J Takala, RAIM and
Complementary Kalman Filtering for GNSS
Reliability Enhancement, IEEE/ION Position,
Location, And Navigation Symposium, pp.948-956,
2006
[11] J Wang, C Xu, & J Wang, Applications of robust
Kalman filtering schemes in GNSS navigation, Int.
Symp. on GPS/GNSS, pp.308-316, 2008
[12] Pratap Misra and Per Enge, Global Positioning
System: Signals, Measurements and Performance,
Ganga-Jamuna Pr, 2006

2235
23rd International Technical Meeting of the Satellite Division of
The Institute of Navigation, Portland, OR, September 21-24, 2010

You might also like