Modern Applied Science; Vol. 11, No.

1; 2017
ISSN 1913-1844 E-ISSN 1913-1852
Published by Canadian Center of Science and Education

GPS/INS/Odometer Data Fusion for Land Vehicle Localization in

GPS Denied Environment
Mohammed Aftatah1, Abdelkabir Lahrech2, Abdelouahed Abounada1 & Aziz Soulhi3
1
Faculty of Sciences and Techniques, Sultan Moulay Slimane University, Beni Mellal, Morocco
2
Multidisciplinary Faculty, Hassan First University, Khouribga, Morocco
3
Superior National School of Mines, Mohamed Fifth University, Rabat, Morocco
Correspondence: Mohammed Aftatah, Faculty of Sciences and Techniques, Sultan Moulay Slimane University,
Beni Mellal, Morocco. Tel: 212-0633-249-019. E-mail: [email protected]

Received: July 29, 2016 Accepted: September 6, 2016 Online Published: October 11, 2016
doi:10.5539/mas.v11n1p62 URL: http://dx.doi.org/10.5539/mas.v11n1p62

Abstract
The main purpose of this paper is to present a fusion approach to bridge the period of Global Positioning System
(GPS) outages using two proprioceptive sensors that are the Inertial Navigation System (INS) and the odometer
in order to assure a continuous localization for land vehicle in urban areas where GPS signal blockage is very
often. Odometer and GPS measures are exploited to correct inertial sensor errors. In fact, during GPS availability,
INS is integrated with GPS to provide accurate localization solution; whereas during GPS outages, the odometer
measurements are used to correct the INS error thereby improving the positioning accuracy and assuring the
continuity of navigation solution. The problem of estimation of vehicle localization is realized by Kalman Filter
(KF) that merges sensor measurements. The paper thus introduces results from simulation and real data.
Keywords: kalman filtering, Inertial Navigation System, odometer, data fusion, GPS, continuous navigation
1. Introduction
Satellite navigation systems such as GPS or Galileo are increasingly popular and used in strategic as well as
public applications. If the safety of persons or property is involved, like in civil aviation, the GPS accuracy is
insufficient to be certified as primary means of navigation. One of the solutions is to integrate the GPS with
other sensors that provide a good availability such as INS and odometry. Integrated navigation systems have
been used intensively in many domains such as aeronautics. However, their application to the automotive
industry shows a big increase due to the low cost inertial measurement units (IMU). INS is a system that delivers
the position, velocity, and attitude of a vehicle by exploiting the output of inertial sensors. The measurements of
the inertial sensors are affected by errors due to physical limitations. The accumulation of these errors leads to
the decrease of the accuracy of navigation solution. Therefore, if the error is not compensated with additional
sensors, the navigation solution of INS can only be trusted during a short period of time (Seo & Lee, 2005).
Nowadays, GPS is very often used as an aiding source to the INS, and the GPS/INS integrated system with good
GPS availability provides more precise dynamic positioning than a stand-alone GPS or INS. However, GPS has
a low sampling rate thus the satellite signals are not always available due to high buildings, tunnels and
mountains, multi-path reflections and bad weather conditions (Noureldin, El-Shafie & Taha, 2007) (Farrell,
1998). To achieve a high level of positioning estimation in an urban area, a new developed integration navigation
system is detailed in this paper. Our proposed method is based on the use of an additional aiding source. This
additional sensor is the odometer. All modern cars are equipped with the Antilock Braking Systems (ABS) that
includes an odometer which measures wheel angular rates and estimate the traveled distance. It is an alternative
to satellite-based navigation technology when the GPS measures are not available. The purpose of our work is to
propose a global modeling of the GPS/INS/Odometer fusion problem with long GPS outages. The estimation of
dynamic characteristics of the vehicle is solved by a KF that fuses the measurements of each sensor to estimate
position, velocity and attitude of the vehicle.
This paper traits the performance of the proposed method in terms of accuracy using both simulated and real data.
For real tests we have used a Novatel GPS receiver and we estimated the INS and measurements from the real
trajectory. In this work additional aiding source information is also considered as an aiding to the positions

62
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

delivered by GPS receiver. This extra information is estimated by an odometer. There are four different
categories of integration approaches: un-coupled (Titterton & Weston, 1996), loosely coupled (LC) (Wolf,
Eissfeller & Hein, 1997) (Solimeno, 2007), tightly coupled (TC) (Li et al., 2006) (Petovello, 2003), and
ultra-tightly coupled (UTC) techniques (Babu & Wang, 2007) (Pany, Kaniuth & Eissfeller, 2005) (Petovello &
Lachapelle, 2006). The second approach uses GPS and odometer position measurements in addition to the
heading angle estimated by the odometery in a Kalman filter that models INS error dynamics. In this paper we
have implemented a Loosely Coupled algorithm that has the ability to exploit additional information when
available. In such a case, a Kalman filter that uses the position and the heading angle, estimated by the odometer,
as measurement has been designed in order to readjust the errors affecting the inertial sensors when GPS signals
are absent.
The rest of this paper is organized as follows: Section 2 discusses the INS and the odometer sensor modeling.
Section 3 describes the KF fusion algorithm and detailed the system model and the observation model. Section 4
offers simulation results. Finally, section 5 is dedicated to conclusions.
2. Sensors Modeling
2.1 INS Modeling
The equations modeling the INS sensors are used to obtain navigation solutions (position, velocity and
orientation of the mobile) from the inertial sensors measurements, i.e. specific forces and angular rates. In this
work these equations are expressed in the ENU frame (East, North and Up). In the following the ENU coordinate
system will be considered as the navigation frame (n-frame). The body frame (b-frame) is defined at the INS
centre. The dynamic equations in the n-frame are given by (Jekeli, 2001) (Abd Rabbou & El-Rabbany, 2015):
r LLa = DV n (1)
v = Cb f ib − (2ωie + ωen ) × v + g
n n b n n n n
(2)

 n = Cn (S(ωb ))
C (3)
b b nb

Where r n = [ φ λ ϕ] is the position in the navigation frame, expressed as latitude, longitude and altitude,
T

vn = [ ve v u ] is the linear velocity, f ibb is the specific force measured by accelerometers expressed in the
T
vn

b-frame, Cnb is a function of the three orientation angles roll φ, pitch θ and yaw ψ and it’s the transformation
matrix from the b-frame to the n-frame and vice-versa for Cbn . ωbnb is the vector of angular velocities about the
axes of the INS provided directly by gyroscopes after removing ωieb and ωenb , ωien is the Earth turn rate in the
navigation frame, ωenn is the turn rate of the n-frame with the respect to the Earth, g n is the local gravity vector,
D is the transformation matrix from the ENU frame to the LLa frame, it’s given by (Titterton & Weston, 2004):

 1 
 0 0
 Rm + h 
 1 
D= 0 0 (1a)
(R
 n + h) cos φ 
 0 0 1
 
 
Where
a(1 − e 2 )
Rm = (1b)
(1 − e 2 sin 2 φ)3/2
a
Rn = (1c)
1 − e 2 sin 2 φ
R m and R n are the radii of curvature in the meridian and prime vertical, respectively (Tang et al., 2005). e is the
eccentricity of the Earth, a and b are the semi-major axis and semi-minor axis of the Earth, respectively.

63
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

(1) Can be expressed in the ENU frame as:

P n = V n (4)
Where P = [ Pe Pn Pu ] is the position vector in the navigation frame.
T
n

2.2 INS error equations

Eventually, the INS error equations are obtained by perturbing the kinematic equations, i.e. (1-3). These error
equations are used in the construction of the GPS/INS/Odometer system model. The perturbation of the position
in (1), velocity in (2) and attitude in (3) can be expressed as (5-7), (Angrisano, 2010):
δr n = Frr δr n + Frv δv n (5)
Where
 −vn 
 0 0 
 (R m + h) 2 
 v e sin φ − ve 
Frr =  0  (5a)
 n + h) cos φ
(R (R n + h) cos φ 
2

 0 0 0 
 
 
Frv = D (5b)
δv = Fvr δr + Fvv δv + Fvε ε + Cb δf ib
n n b
(6)
Where
 ve vn v e v u − v e v n tan φ 
 2ωe (v u sin φ + v n cos φ) + 0 
 (R n + h) cos 2 φ (R m + h) 2 
 
(6a)
v e2 v n v u + v e tan φ
2

Fvr =  −2ωe v e cos φ − 0 

 (R n + h) cos 2 φ (R m + h) 2 
 − ve2
vn2
2g 
 −2ωe v e sin φ 0 − + 
 (R n + h) (R m + h) 2 R e + h 


R e is the radius of the earth and g is the gravity,

 v n tan φ vu v e tan φ ve 
 − 2ωe sin φ + −2ωe cos φ − 
 R m
+ h R n
+h Rn + h Rn + h 
 2v e tan φ −vu −vn  (6b)
Fvv =  −2ωe sin φ − 
 Rn + h Rm + h Rm + h 
 2v e vn 
 2ωe cos φ + 0 
 R +h Rm + h 
 n 

 0 fu −f n 
 
Fvε =  −f u 0 fe  (6c)
 f −f e 0 
 n
f ibn = [ f e f n f u ] is the specific force in the ENU frame,
T

ε = Fεr δr n + Fεv δv n + Fεε ε + Cnb δωibb (7)

Where

64
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

 −vn 
 0 0 
 (R m + h) 2 
 ve 
Fεr =  ωe sin φ 0  (7a)
 (R n + h) 
2

 ve v e tan φ 
 −ωe cos φ − 0 
 (R n + h) cos 2 φ (R n + h) 2 

 1 
 0 0
 (R m + h) 
 −1 
Fεv =  0 0 (7b)
 Rn + h 
 − tan φ 
 0 0 
 Rm + h 
 v e tan φ ve 
 0 ωe sin φ + −ωe cos φ − 
 Rn + h Rn + h 
 v tan φ −vn 
Fεε =  −ωe sin φ − e 0  (7c)
 Rn + h Rm + h 
 ve vn 
 ωe cos φ + 0 
 R +h Rm + h 
 n 
The INS mechanization in the navigation frame is summarized in figure 1 as follows:

n
g
rn

V
n n
b V Pn
f ib
C
n
b

vn
− (2ω + ω )
n
ie
n
en
ϕ, θ, ψ

C
n
ω
b

ω
b nb n b
ib C b
−ω
n
en

−ω
b
in
C n
−ω
n
in

−ω
n
en

Figure 1. INS mechanization in ENU frame, adopted from (Schultz, 2006) (Quinchia et al., 2013)

2.3 Odometer Modeling

The kinematic equations of the char vehicle model figured in figure 2 are given by (Abuhadrous, 2005):

  ωR + ω L 
 u = rwheel  
  2 
 (8)
θ = r  ωR − ωL 
wheel  
  e 

Where ωR and ωL are the angular velocities of the right and left wheels, respectively. θ is the heading angle rate

65
mas.ccsenett.org Modernn Applied Sciencce Vol. 11, No. 1; 2017

(angular vvelocity) of thee vehicle in thee x-y plane, rwhheel is the radiuus of the two w
wheels, e is thee length of the
e axle
of the vehhicle (distance between the two wheels), v R = rwheel ωR annd v L = rwheel ωL aare the velocitties of the leftt and
right wheeels respectivelyy, u is the aveerage of the two velocities.
The angulaar velocities or the measuredd displacemennts in (8) for thhe right wheel are written as (Lahrech, Bou
ucher
& Noyer, 22005):

ωRt+1 = rR−1 (ωRt+,x1 ) 2 + (ωRt +,y1 ) 2 (9)

Where

e v ty+1 v ty 
ωRt +,x1 = v xt Δt +  −  (9a)
2  (v xt+1 ) 2 + (v ty+1 ) 2 (v xt ) 2 + (v ty ) 2 


e v xt+1 v xt 
ωRt +,y1 = v ty Δt −  −  (9b)
2  (v xt+1 ) 2 + (v ty+1 ) 2 (v xt ) 2 + (v ty ) 2 

And for thhe left wheel (L
Lahrech, Boucher & Noyer, 22005):

ωLt+1 = rL−1 (ωLt+,x1 ) 2 + (ωLt +,y1 ) 2 (10)

Where

e v ty+1 v ty 
ωLt +,x1 = v xt Δt −  −  (10a)
(
2  (v xt +1 ) 2 + (v ty+1 ) 2 (v xt ) 2 + (vv ty ) 2 


e v xt+1 v xt 
ωLt +,y1 = v ty Δt +  −  (10b)
(
2  (v xt+1 ) 2 + (v ty+1 ) 2 (v xt ) 2 + (vv ty ) 2 


rR and rL are the radii of

o the right annd left wheels respectively, e is the distan
ance separatingg the two poin
nts of

contact of the wheels wiith the ground.

Figuure 2. The charr vehicle modeel, kindly takenn from (Abuhaadrous, 2005)

The odommeter measurem ments are explloited in case of GPS outagees. The discreete form of thee equations used to
obtain possition from odoometer measurres can be exprressed as (Abuuhadrous, 20055):

66
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

 x k +1 = x k + (u cos θk +1 )Δt

 y k +1 = y k + (u sin θk +1 )Δt (11)
 
θk +1 = θk + θΔt
Where x k and y k denote the position in the center of the axis, Δt is the sampling time.
3. Kalman Filter Fusion Modeling
A KF is chosen to fuse the measurements of the INS and GPS when sufficient satellite signals are available and
to integrate INS measures with odometer in the opposite case. The filter structure is shown in figure 3. The
Kalman filter algorithm has two main steps: the first consists in predicting the state based on the system model
and the second is dedicated to update the state based on the measurements (Grewal & Andrews, 1993).

x INS , y INS , z INS

x GPS , y GPS , z GPS

x INS , y INS , ψ INS

x Odo , y Odo , ψ Odo

Figure 3. The filter structure

3.1 System Model

The INS error equations are used in the Kalman filter as the system error dynamic model of integrated
navigation (Angrisano, 2010):

 δr n   F Frv 03 03 03   δr n 
   rr  
 δv n   Fvr Fvv Fvε Cnb 03   δv n 
 =  ε  
C nb   ε 
X   =  Fεr Fεv Fεε 03
 
 δb   03 03 03 α bacc 03   δb acc 
 acc 
 δb   03 03 03 03 α bgyro   δb gyro 
 gyro  (12)
 03 03 03 03 
 n   ηacc 
 Cb 03 03 03   
 ηgyro 

+ 03 Cbn
03 
03 
  η 
 03 03 I3 03   bacc 
η 
0 03 03 I3   bgyro 
 3
T
Where X = δr n δv n ε δbacc δbgyro  is the state vector, δr n is the position errors, δv n is the linear
T T
velocity errors, ε is the Euler angles errors, δbacc = δbaccx δb accy δb accz  and δbgyro = δbgyrox δbgyroy δbgyroz 
are the biases of accelerometers and gyroscopes, 0 3 and I3 are zero and identity matrices of size 3, the drift of
these biases can be modeled as a first-order Gauss-Markov process (Hou, 2004) represented as follow:
−1
δb acci = δb acci + ηbacci (12a)
τbacci

67
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

−1
δb gyroi = δb gyroi + ηbgyroi (12b)
τbgyroi
Where i=x, y, z,
−1 T
=  τbaccx τbaccy τbaccz  is the correlation times for the accelerometers,
α acc 

−1 T
=  τbgyrox τbgyroy τbgyroz  is the correlation times for the gyroscopes,
α gyro 

T
ηacc = ηaccx ηaccy ηaccz  is the accelerometers noises,

T
ηgyro = ηgyrox ηgyroy ηgyroz  is the gyroscopes noises,

T T
ηbacc = ηbaccx ηbaccy ηbaccz  and ηbgyro = ηbgyrox ηbgyroy ηbgyroz  are the Gauss-Markov process driving noises,

Because the values of position error in (9) are very small, since they are expressed in radian, and can cause
numerical instability inside the filter, to avoid this problem it is better to express the position error in the ENU
frame (East, North and Up coordinates) (Shin, 2001), the new dynamic model is :

 δP n   F FPv 03 03 03   δP n 
   PP  
 δv n   FvP Fvv Fvε Cbn
03   δv n 
 =  ε  
C bn   ε 
X   =  FεP Fεv Fεε 03
 
 δb   03 03 03 α bacc 03   δb acc 
 acc 
 δb   03 03 03 03 α bgyro   δb gyro 
 gyro  (13)
 03 03 03 03 
 n   ηacc 
 Cb 03 03 03   
 ηgyro 

+ 03 Cbn
03 
03 
  η 
 03 03 I3 03   bacc 
η 
0 03 03 I3   bgyro 
 3
Where
 v e tan φ − ve 
0 
 Rm + h Rn + h 
 −vn 
FPP =  0 0  (13a)
 Rm + h 
 
0 0 0 
 
 

1 0 0
FPv = 0 1 0 (13b)
0 0 1 

68
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

 2ωe (v u sin φ + v n cos φ) ve vn v e (v u − v n tan φ)

0 + 
 Rm + h (R m + h)(R n + h) cos 2 φ (R n + h) 2 
 −2ωe v e cos φ v e2 −2ωe v e cos φ v e2 
FvP =  0 − −  (13c)
 Rm + h (R m + h)(R n + h) cos 2 φ Rm + h (R m + h)(R n + h) cos φ 
2

 −2ωe v e sin φ − v 2n v e2 2g 
0 − + 
 Rm + h (R m + h) 2 (R n + h) 2 R e + h 
 

 −vn 
0 0 
 (R m + h) 2 
 ωe sin φ ve 
FεP =  0  (13d)
 Rm + h (R n + h) 
2

 −ωe cos φ ve v e tan φ 

0 − 
 R + h (R + h)(R + h) cos 2 φ (R n + h) 2 
 m m n

The discrete form of the system model can be written as:

 I3 + FPP Δt FPv Δt 03 03 03 
 
 FvP Δt I3 + Fvv Δt Fvε Δt C bn Δt 03 
X k =  FεP Δt Fεv Δt I3 + Fεε Δt 03 Cb Δt
n
 X k−1
 
 03 03 03 I3 + α bacc Δt 03 
0 (14)
 3 03 03 03 I3 + α bgyro Δt k −1
 03 03 03 03 
 n   ηacc 
 C b 03 03 03   
ηgyro 
+  03 C nb 03 03  
  η 
 03 03 I3 03   bacc 
η 
0 0 03 I3 k −1  bgyro k −1
 3 3

Where Δt is the sampling time.

3.2 Observation Model
3.2.1 GPS Measurements
When the GPS signals are available, position and velocity solution from GPS are integrated with INS at the rate
of 1 Hz. The measurement model for GPS/INS loosely coupled scheme is:

 x INS − x GPS 
ZGPS = H GPS .X + υ GPS
=  y INS − yGPS  + υGPS (15)
 z INS − z GPS 

Where υ  ℵ(0, R GPS ) is a white Gaussian noise.

GPS

The observation matrix is:

1 0 0 01x ( n −3) 
 
H GPS = 0 1 0 01x ( n −3)  (16)
0 0 1 01x ( n −3) 


Where n is the size of the state vector, I3 x 3 is the identity matrix, 03x 3 is a zero matrix and the subscript (m × n)

shows the number of matrix rows (m) and columns (n) . GPS measurements have as covariance matrix:

69
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

σ 2xGPS 0 0 
 
R GPS = 0 σ yGPS
2
0  (17)
0 0 σ 2zGPS 

Where σ 2PGPS is the covariance of PGPS and P=x, y, z.
3.2.2 Odometer Measurements
In denied environments when GPS signals are not available, position solution from the odometer is integrated
with the INS at the rate of 1 Hz. In this case, the measurement model is given by:

 x INS − x Odo 
ZOdo = H Odo .X + υOdo =  y INS − yOdo  + υOdo (18)
ψ INS − ψ Odo 

Where υ  ℵ(0, R Odo ) is a white Gaussian noise.

Odo

The observation matrix has the following form:

1 0 01x 6 0 01x 6 
H Odo = 0 1 01x 6 0 01x 6  (19)
0 0 01x 6 1 01x 6 

Where n is the size of the state vector.

The covariance matrix of odometer measurements is given by:

 σ 2xOdo 0 0 
 
R Odo = 0 σ yOdo
2
0  (20)
0 0 σ ψ2 Odo 

Where σ 2POdo is the covariance of POdo and P=x, y, z.
4. Simulation Results
The section below is dedicated to present some results from our fusion algorithm applied to simulation data in
the purpose to quantify the contribution of the developed solution in terms of accuracy and availability of vehicle
positioning in urban areas.
4.1 Simulated Scenario
In the simulated trajectory of the vehicle, during 26 minutes and 37 seconds (figure 4), there are three zones
where the GPS signals are not available mainly because of the masking of satellite signals or an insufficient
number of visible satellites: the three zones have duration of 70 seconds each. During these periods, the
estimator uses the odometer measures in order to correct inertial sensor errors.

70
mas.ccsenett.org Modernn Applied Sciencce Vol. 11, No. 1; 2017

5
500

4
450

4
400
GPS outage 3
GPS outag
ge 2
3
350

3
300

North (meters)
2
250

2
200

1
150

utage 1
GPS ou
1
100

50

0
-200 -
-150 -10
00 -50 0 50
East (meters)

Figgure 4. Trajecttory taken for simulation inccluding long G

GPS outages

4.2 Vehiclee Localization

Here we ppresent the resuults from our fusion algorithhm applied to simulation daata. During thee GPS outagess, the
filter uses the odometer measures to ccorrect the errrors affecting tthe inertial sennsors. To calcuulate the errorrs the
reference ttrajectory of thhe vehicle is ussed.

0.1 0.1
ve error ve error
GPS outages GPS outages
0.08 vn error 0.08 vn error
vu error vu error
0.06 0.06

0.04 0.04
Velocity error (meters/second)

Velocity error (meters/second)

0.02 0.02

0 0

-0.02 -0.02

-0.04 -0.04

-0.06 -0.06

-0.08 -0.08

-0.1 -0.1
0 200 400 600
0 800 1000 1200 1400 1600 0 200 400 600 800 1000
0 1200 1400 1600
Time (seconds) Time (seconds)

Fiigure 5. Velociity error of the vehicle Figure 6. Posiitioning error oof the vehicle

500
475
450

Reference 474.5
400 Estimated

350 474

300
North (meters)
North (meters)

473.5

250

473
200

150 472.5

100
472

50
471.5
0
-200 -150 -100 -50 0 50 -26.6 -26.4 -26.2 -26 -25.8 -25.6 -25.4 -25.2 -25
5
East (meters) East (meters)

Figuure 7. Estimateed trajectory of the vehicle Figuree 8. Zoom on rreference and estimated traje
ectories

Figure 5 shows the veelocity error of the vehiclle. We can notice that this error is approximately 0.005
0

71
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

meters/second in the presence of GPS signal blockages. The use of an additional aiding source, the odometer,
leads to good results in terms of vehicle velocity estimation. In fact, the estimated error of the vehicle velocity is
significantly reduced during GPS signal outages. As shown in figure 6, the proposed method improves
significantly the vehicle positioning accuracy even during long GPS outages: the error does not exceed 0.3
meters for the horizontal component and the East and North components. A 2-D plot of the simulated trajectory
and the estimated one are show below. It is clear by observing Fig. 7 and Fig. 8 that the position errors of the
GPS/INS/Odometer integrated navigation system are smaller than those of GPS/INS integrated navigation
system. Especially, where GPS signal is absent. From these results, it can be seen that the proposed algorithm
fusion improves the accuracy of vehicle localization. In addition to this, the GPS/INS/Odometer integrated
system realized by using the proposed data fusion methodology leads to good performance compared with the
GPS/INS integrated system during GPS outages.
4.3 Real Data Test Results
This section is dedicated to the presentation of the results obtained by the application of the new method to a real
trajectory.
4.3.1 Urban Transport Scenario
Tests were conducted along a track located within Calais city, France. Our test vehicle is equipped with a
Novatel GPS receiver that calculates the position. The measurement campaign was conducted in an urban area
during 5 minutes and 25 seconds. There are two areas where GPS positioning is absent due to an insufficient
number of satellites: the first has duration of 40 seconds and the second 31 seconds. During these periods, the
estimator uses odometer measurements which are delivered at a frequency of 1 Hz, as the GPS measurement
when available. The test track is shown in figure 9.
6
x 10
5.6458
GPS available
GPS outage
5.6457

5.6456

5.6455
North(meters)

5.6454

5.6453

5.6452

5.6451

5.645
4.2 4.202 4.204 4.206 4.208 4.21 4.212 4.214 4.216 4.218
East (meters) 5
x 10

Figure 9. Real trajectory of the vehicle with long GPS outages

4.3.2 Vehicle Positioning

It is clear from figure 10 that the performance of the proposed method (gray rectangles) improves significantly
the performance of the GPS/INS/Odometer integrated system with GPS outages (the rest of figure 10) in terms
of velocity estimation. In fact, this method trusts the odometer measurements (East and North positions in
addition to the heading angle) when GPS measures are not available and permits to readjust the INS navigation
solution.

72
mas.ccsenett.org Modernn Applied Sciencce Vol. 11, No. 1; 2017

8
3
Pe error
Pn error
ve error GPS outages GPS outages
vn error 6 Pu error
2
vu error

1
Velocity error (meters/second)

Position error (meters)

2

-1
-2

-2
-4

-3 -6
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350
Time (seconds) Time (seconds)

F
Figure 10. Veloocity error of thhe vehicle Figure 11. Poositioning erroor of the vehicle

6
6 x 10
x 10
5.6458
5.6451 Reference
Reference
Estimated
Estimated
Estimated by INS
5.6457 Estimated by INS
S
5.6451

5.6456 5.6451

5.6451
5.6455
North (meters)
North (meters)

5.6451
5.6454

5.6451

5.6453
5.6451

5.6452
5.6451

5.6451 5.6451

5.6451
5.645
4.2 4.202 4.204 4.206 4.208 4.21 4.212 4.214 4.216 4.218
East (meters) 5
4.2
2102 4.2104 4.2106 4.2108 4.211 4.2112 4.2114 4.2116 4.2118 4.212 4.2
2122
x 10 East (meters) 5
x 10

F
Figure 12. The generated trajectory from Figure 13. Zoom on GPS S/INS/Odometer
G meter integratedd navigation
GPS/INS/Odom llocalization solution
system

Figure 11 shows the threee-dimensionaal position estim mation during GPS outages (gray rectangles) and in pressence
of GPS positioning (thee rest of figuure 11). As caan be depictedd from figure 11, the proposed method error
estimates qquickly converrge to stable values after thee initial transitiion period. Thee error doesn’tt exceed 1.5 meters
m
for East aand North coomponents. M Moreover, the points duringg GPS outagees are selecteed to evaluate e the
localizatioon accuracy off the proposed method. The table below prresents the staandard deviatioon of position error
for both simulation dataa and real sceenario during G GPS outages, respectively. The table 1 shows that the new
solution reesults are obviiously better thhan the INS standalone systeem where GPS S signals are aabsent; the stan
ndard
deviation of the combined solution is 0.0217 meeters and 0.2597 meters forr both simulaation and real data
respectivelly. However, forf standalone INS the standdard deviation becomes 0.14437 meters andd 1.1487 meters by
the end off the vehicle traajectory.

Table 1. Poosition error sttandard deviatiion in meters

Data GPS outaages
INS
S only IN
NS and odometter
Simulattion 0.1437 0.02177
Real 1.1487 0.25977

73
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

5. Conclusion
In summary, this paper proposes a new technique for GPS/INS/Odometer integration based on Kalman filter that
minimizes the INS error and provides continuous estimation of vehicle position, velocity and attitude. This
combination permit to avoid disadvantages of a stand-alone sensor in order to establish long-term navigation in
GPS denied environments. The proposed method is a loosely-coupled solution that fuses the available measures.
As described earlier the system presented in our work consists of the GPS receiver, an INS, and a vehicle
odometer. The new technique solution was tested with real and simulated trajectories. The results for both
trajectories were presented. The first trajectory contains 3 simulated GPS outages of 40 seconds duration were
introduced intentionally. This was repeated four different. The proposed solution based on KF achieves
centimeter-level positioning accuracy. The second trajectory is an urban scenario with natural GPS absence
because of satellites signal blockage, where the proposed system solution was tested to demonstrate its
performance in such environments.
In this work we have examined the odometer bridging performance for the Inertial Navigation System during
GPS outages, and the results obtained from simulation and real tests confirm that using an additional aiding
source, the odometer, provides more accurate navigation estimation than GPS/INS integrated system can offer in
urban areas.
The proposed GPS/INS/Odometer integrated navigation system overcomes the problems due to environment of
GPS signals propagation and shows good performance for land vehicle navigation solution using inertial sensors
and an odometer in urban environment where the blockage of GPS satellites signals is very often.
References
Seo, J., & Lee, J. G. (2005). Application of nonlinear smoothing to integrated GPS/INS navigation system.
Journal of Global Positioning Systems, 4(2), 88–94. http://dx.doi.org/10.5081/jgps.4.1.88
Noureldin, A., El-Shafie, A., & Taha, M. R. (2007). Optimizing neuro-fuzzy modules for data fusion of vehicular
navigation systems using temporal cross-validation. Engineering Applications of Artificial Intelligence, 20,
pp. 49–61. http://dx.doi.org/10.1016/j.engappai.2006.03.002
Farrell, J. (1998). The Global Positioning System and Inertial Navigation. McGraw-Hill Professional, New York.
Titterton, D. H., & Weston, J. L. (1996). Strapdown Inertial Navigation Technology (2nd Ed.). The American
Institute of Aeronautics and Astronautics: Reston, VA, USA.
Wolf, R., Eissfeller, B., & Hein, G. (1997). A Kalman Filter for the Integration of a Low Cost INS and an attitude
GPS. In Proceedings of the International Symposium on Kinematic Systems in Geodesy, Geomatics and
Navigation, Banff, AB, Canada, pp. 143–150.
Solimeno, A. (2007). Low-Cost INS/GPS Data Fusion with Extended Kalman Filter for Airborne Applications.
Master Thesis, Universidad Tecnica de Lisboa, Lisboa, Portugal.
Li, L., Wang, Y., Rizos, C., Mumford, P., & Ding, W. (2006). Low-Cost Tightly Coupled GPS/INS Integration
Based on a Nonlinear Kalman Filtering Design. In Proceedings of the 2006 National Technical Meeting of
The Institute of Navigation, Monterey, CA, USA, pp. 958–966.
Petovello, M. (2003). Real-time Integration of a Tactical-Grade IMU and GPS for High-Accuracy Positioning
and Navigation. Ph.D. Thesis, Department of Geomatics Engineering, University of Calgary, Calgary, AB,
Canada.
Babu, R., & Wang, J. (2007). Real-Time Data Analysis of Ultra-Tight GPS/INS. In Proceedings of IGNSS
Symposium 2007, Sydney, NSW, Australia.
Pany, T., Kaniuth, R., & Eissfeller, B. (2005). Deep Integration of Navigation Solution and Signal Processing. In
Proceedings of the 18th International Technical Meeting of the Satellite Division of The Institute of
Navigation (ION GNSS 2005), Long Beach, CA, USA, pp. 1095–1102.
Petovello, M. G., & Lachapelle, G. (2006). Comparison of Vector-Based Software Receiver Implementations
with Application to Ultra-Tight GPS/INS Integration. In Proceedings of the 19th International Technical
Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2006), Fort Worth, TX, USA,
pp. 1790–1799.
Jekeli, C. (2001). Inertial navigation systems with geodetic applications. Walter de Gruyter: Berlin, Germany.
Abd R. M., & El-Rabbany, A. (2015). Integration of GPS Precise Point Positioning and MEMS-Based INS
Using Unscented Particle Filter. Sensors, 15(4), 7228-7245. http://dx.doi.org/10.3390/s150407228

74
mas.ccsenet.org Modern Applied Science Vol. 11, No. 1; 2017

Titterton, D., & Weston, J. L. (2004). Strapdown Inertial Navigation Technology (2nd Ed.). The American
Institute of Aeronautics and Astronautics and the institution of electrical engineers: Reston, FL, USA.
Tang, J., Chen, Y., Niu, X., Wang, L., Chen, L., Liu, J., … Hyyppä, J. (2015). LiDAR Scan Matching Aided
Inertial Navigation System in GNSS-Denied Environments. Sensors, 15 (7), 16710-16728.
http://dx.doi.org/10.3390/s150716710
Angrisano, A. (2010). GNSS/INS integration methods. PhD thesis, Research in Geodetic Sciences and
Topographic XXVIII Cycle, Department of Applied Sciences, University of Naples, Naples, Italy.
Schultz, C. E. (2006). INS and GPS Integration. M.Sc. Thesis, Technical University of Denmark, Kongens
Lyngby, Denmark.
Quinchia, A. G., Falco, G., Falletti, E., Dovis, F., & Ferrer, C. (2013). A Comparison between Different Error
Modeling of MEMS Applied to GPS/INS Integrated Systems. Sensors, 13(8), 9549-9588.
http://dx.doi.org/10.3390/s130809549
Abuhadrous, I. (2005). Onboard real time system for 3D localization and modelling by multi-sensor data fusion.
domain other. Ecole Nationale Superieure des Mines de Paris, 62-64.
Lahrech, A., Boucher, C., & Noyer, J. C. (2005). Accurate vehicle positioning in urban areas. Conference of the
IEEE Industrial Electronics Society, 31. http://dx.doi.org/10.1109/iecon.2005.1568953
Grewal, M. S., & Andrews, A. P. (1993). Kalman Filtering: Theory and Practice. Prentice-Hall, Inc.: Hoboken,
NJ, USA.
Hou, H. (2004). Modeling Inertial Sensors Errors Using Allan Variance. The University of Galgary: Calgary, AB,
Canada. UCGE Reports, No. 20201.
Shin, E. (2001). Accuracy Improvement of Low Cost INS/GPS for Land Application. MSc Thesis, Department of
Geomatics Engineering, University of Calgary, Canada. UCGE Report, No. 20156.

Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution
license (http://creativecommons.org/licenses/by/4.0/).

75