International Conference on Engineering, Science and Advanced Technology (ICESAT-2023), Technical Engineering

College / Mosul, Iraq

On the Location Accuracy of the Multi-antenna GPS

Ziyad Khalaf Farej PhD Ahmed Nadeem Ismail
Department of Computer Technology Engineering Department of Computer Technology Engineering
Technical Engineering College Northern Technical University Technical Engineering College Northern Technical University
Mosul, Iraq Mosul, Iraq
drziyad.farej@ntu.edu.iq ahmadna1985@ntu.edu.iq

Abstract—Despite the significant improvements in signal Although systems like smart cities, driverless vehicles,
processing techniques currently in use, GPS receivers still vehicle ad hoc networks, and drones will significantly boost
encounter significant obstacles, like multipath, that continues to the need for radio communication services in the near future
be a primary source of range inaccuracy. GPS multipath affects [3], the multipath remains a major causes of location mistakes
the acquisition and tracking components inside the receiver,
resulting in inaccurate readings or a loss of GPS satellite signal
while utilizing GPS and other Global Navigation Satellite
lock. As a result, accurate location and velocity cannot be Systems. Multipath signals can seriously impair GPS
obtained from GPS devices. The main objective of this research pseudorange measurements because they alter the form of the
is to enhance the performance of the GPS system by reducing correlation function that is used to estimate time delays [4].
the effect of multipath on the accuracy of vehicle’s location
determination. In this paper, a vehicle tracking system that is
In this research paper a multi-antenna GPS is designed and
based on multi antenna sensors module is proposed to increase implemented to decrease the effect of multipath problem and
the accuracy of vehicle location. The proposed system uses the compare its performance with single antenna GPS. The
Arduino Mega board as microcontrollers and low-cost global general configuration of the proposed system, which consists
positioning system sensor. Under the suggested microcontroller, of three antenna GPS sensors type Neo-6m with Arduino
the performance of the proposed GPS system in determining the Mega microcontroller is connected to a laptop. The three
precision of vehicle location, is investigated for different speeds sensors are connected so that 120 degrees coverage is obtained
at different locations. The minimum and maximum location for every one of them and the space between every two
accuracy results are 2.69 and 31.39 meters at speeds of 0 and 100
adjacent antennas is 10 centimeters as shown in Fig. 1. So, 360
Km/h respectively. The findings displays that the proposed
system can operate in any weather and offer more precise real-
degrees converge can be made for the three sensors.
time object position.
10 cm
Keywords—GPS, Neo-6M, Multipath

I. INTRODUCTION
The use of location tracking systems for implementation
in a variety of applications, including but not limited to sports,
health, and vehicle detection and monitoring applications, has
seen a rise in popularity. These applications cover the civilian
and military sectors, and each one has specific criteria for size,
price, and accuracy. The location identification and tracking
of dementia patients who experience memory loss issues is an
example of a health application [1].
High-quality radio communication services will be needed
in the near future due to the massive expansion in utilizing of
real-time locating devices. In real-world applications,
scattering and reflection brought on by structures like trees,
buildings, and land vehicles reduce GPS performance.
Buildings in metropolitan environments frequently reflect
or block GPS signals, resulting in multipath effects and non-
line of sight (NLOS) reception, respectively, which reduce
GPS location accuracy. While a better receiver design can
somewhat lessen the multipath impact, it is unable to handle
NLOS. Measurement-based modeling techniques have
showed promise in minimizing the impact of NLOS signal Fig. 1. The proposed 3 sensors-GPS with 120 degrees coverage for each
reception [2]. antenna sensor

29
International Conference on Engineering, Science and Advanced Technology (ICESAT-2023), Technical Engineering
College / Mosul, Iraq

II. RELATED WORKS position as the optimum distribution for describing various
Many researchers have considered the problem of circumstances on the three GPS frequencies that are currently
multipath and its effect on the accuracy of the location accessible. Then, for scenarios involving urban canyons,
systems. R. Sun et al utilized the three characteristics of C/N0, typical urban regions, rural areas, and densely vegetated areas,
pseudorange residuals, and satellite elevation angle with a they show typical values of fading coefficients in L1, L2C,
gradient boosting decision tree (GBDT) based classification and L5 signals. These coefficients may utilized to assess
method to increase the accuracy of signal reception receiver performance in similar scenarios or can be used into
categorization. Several classifiers, including those based on weight measuring techniques to enhance positioning
decision trees, distance-weighted k-nearest neighbors, and the calculation [3].
adaptive network-based fuzzy inference system, are compared III. MULTIPATH PROBLEM AND THE METHOD OF AVOIDANCE
to the proposed technique through experiments (ANFIS). The IN GPS
GBDT-based system obtains a classification accuracy of
100%, 82%, and 86% for LOS, multipath, and NLOS signals, When employing signal antenna GPS, the multipath error
respectively, according to test findings from static receivers in in GPS can be calculated mathematically as [7][8]:
urban environments. With similar outcomes of 100%, 82%, 𝑟(𝑡) = 𝑅 + 𝑐(𝑡 − 𝜏) + 𝛿𝑚(𝑡) + 𝜀(𝑡) (1)
and 84% for the Distance-Weighted KNN, 99%, 70%, and
65% for the ANFIS, and 98%, 35%, and 95% for the where:
conventional decision tree, this outperforms the other three 𝑟(𝑡) is the signal received at time t.
methods. The suggested GBDT with multi-feature-based
approach may offer a positioning accuracy increase of 34.1% 𝑅 represents the actual distance between the GPS receiver and
over the conventional C/N0 based method when it comes to satellite.
NLOS detection and exclusion [2].
𝑐 is light speed
Xi Hong et al suggested an efficient subspace-based
estimator that works in conjunction with the standard delay- 𝜏 is the signal lag brought on by the ionosphere and other
locked loop for the L1 C/A signal of GPS. The structure of the signal lag causes.
receiving array may be used to create a subspace rotation 𝛿𝑚(𝑡) is the multipath error, and it is the result of the GPS
invariance, allowing for the estimation of both the angles of signal reflecting of many objects.
arrival and the relative delays of incident rays. they strength
the LOS ray and create a modified delay tracking output 𝜀(𝑡) is the measurement of noise and errors.
without multipath bias by merging the estimates from DLL The received signal at each antenna may have a distinct
and subspace-based estimators. The researchers showed that multipath error, δm(t), when many antennas are employed. By
the suggested technique has the best multipath mitigation in resolving a set of equations that include the recorded distances
the delay estimate, according to simulation findings, when from each antenna, the GPS receiver can utilize these many
compared to other methods like the narrow early-minus-late observations to calculate the real distance (R). The equations
and the multipath estimating delay lock loop [5]. of the system as follows [9][10]:
James E. et al. provided a vehicle pose estimation method 𝑟1(𝑡) = 𝑅 + 𝑐(𝑡 − 𝜏) + 𝛿𝑚1(𝑡) + 𝜀1(𝑡) (2)
that strongly integrates a low-cost MEMS inertial sensor with
multi-antenna carrier-phase differential GNSS (CDGNSS), 𝑟2(𝑡) = 𝑅 + 𝑐(𝑡 − 𝜏) + 𝛿𝑚2(𝑡) + 𝜀2(𝑡) (3)
vehicle dynamics restrictions, and a MEMS inertial sensor.
Unscented linearization enables utilizing of conventional 𝑟3(𝑡) = 𝑅 + 𝑐(𝑡 − 𝜏) + 𝛿𝑚3(𝑡) + 𝜀3(𝑡) (4)
integer least squares for ambiguity resolution while implicitly Where:
imposing known-baseline-length limitations and taking
advantage of the inter-baseline correlations in the multi- 𝑟1(𝑡) , 𝑟2(𝑡) and 𝑟3(𝑡) are the signals received at (3) different
baseline issue. The effect of conditioning the filter state on antennas at time t.
wrong numbers is reduced by the development of a unique
𝛿𝑚1(𝑡) , 𝛿𝑚2(𝑡) and 𝛿𝑚3(𝑡) are the associated multipath errors
false fix detection and recovery mechanism. The proposed
at time t.
technique achieves a 96.6% and 97.5% integer fix availability
and a twelve cm and ten cm overall (fix and float) 95th 𝜀1(𝑡) , 𝜀2(𝑡) , and 𝜀3(𝑡) are the associated noise error for these
percentile horizontal positioning error when tested on TEX- antennas at time t.
CUP dataset [6].
𝑅 is the real distance.
V. Carvalho et al offered data supports the distribution as
the best method to deal with multifrequency GPS multipath 𝑐 is light speed
channels inside urban, rural, and forest regions. The 𝜏 is the signal lag brought on by the ionosphere and other
Nakagami-m and Rice models are used to analyze the fading signal lag causes.
of GPS signals. To explore propagation scenarios including
scattering clusters and specular reflections, a synthetic signal The GPS receiver can calculate the multipath errors from
simulator was constructed. Moreover, experimental the received signals, 𝑟1 (𝑡), 𝑟2 (𝑡) and 𝑟3 (𝑡). These signals are
measurements are provided to support the - distribution's used to filter out or decrease the influence of multipath on the
GPS location calculation, by solving this set of equations. The

30
International Conference on Engineering, Science and Advanced Technology (ICESAT-2023), Technical Engineering
College / Mosul, Iraq

mathematical equation for multipath error may be used to connected to VCC and GND of the microcontroller too. It has
estimate the correct distance and decrease the effects of high performance 50-channel u-blox 6 core technology
multipath, and employing many antennas in GPS can enhance installed. It has an effective correlation of at least 1,000,000.
the accuracy and reliability of the position computation by Because it uses Super Sense® kick start technology, it can still
decreasing the impacts of multipath [9], [10]. detect a GPS satellite signal by accelerating even when the
signal is poor. The anti-interference capabilities of the U-blox
Figure 2 shows a receiver receives (3) signals, including GPS receiver are a result of its innovative RF design and
direct LOS signal, a signal reflected by a structure, and a cutting-edge noise reduction technology. The u-blox GPS
signal reflected by the ground. The receiver is unable to receiver also contains additional technologies including GPS,
differentiate between LOS and Non-LOS signals, resulting in inertial navigation, and precise timing [12].
the tracking of a combined signal that contains multipath
errors, leading to inaccurate measurements. The three
antennas will deal with this problem and will decrease
multipath in the signal and increase GPS accuracy.

Fig. 3. Circuit connection diagram of the proposed multi-antenna sensors

Fig. 2. Example of multipath GPS

The three antennas used in the proposed system make a

spatial diversity and every one of them covers 120 degrees to
deal and combat the multipath problem.
IV. DESIGN AND IMPLEMENTATION OF PROPOSED GPS
The circuit connection diagram of the proposed multi-
antenna GPS is shown in Fig. 3. Its hardware implementation
as shown in Fig. 4 has three (antennas) GPS sensor type (Neo-
6m) and microcontroller (Arduino Mega AT mega 1280).
These three antennas will reduce multipath problem. The
proposed GPS has been designed and implemented as follows:
1. Hardware part
a) Arduino Mega:
It employs the ATmega1280, a surface mount chip that is
permanently attached to the board, a processor with more
Fig. 4. Hardware of the proposed multi-sensor GPS
input output pins, and boasts a sizable number of input output
ports that are smartly added as additional connections at one
end of the board. Hence, if you accidently damage the CPU,
you cannot replace it, unlike with the Uno and comparable
boards. The additional connections are located at the board's
end. There are 4 serials in it. The Mega has additional features
like (54 I/O pins, 128KB of flash memory and 8KB RAM and
4KB EEPROM) [11].
b) GPS sensor:
The sixth generation of GPS receivers, the neo-6m GPS
module (as shown in Fig. 5), utilized in the design is made by
the Swiss business u-blox. The Tx, Rx pins of every sensor are
connected to Tx, Rx of the microcontroller. Every sensor Fig. 5. Shows the Neo-6m GPS sensor (antenna module)

31
International Conference on Engineering, Science and Advanced Technology (ICESAT-2023), Technical Engineering
College / Mosul, Iraq

2. Software part 100 36.38075256 43.16581726 21.48 36.3807411 43.16582107 21.99

36.38073730 43.165824890 22.32
a) Arduino IDE
The application known as Integrated Development
Table II shows the errors in the second point coordinates
Environment, or IDE, was made available by Arduino and is
and also for (5) speeds on the same point. In speed of (0) km,
primarily utilized for editing, compiling, and uploading code
the error range changes between (2.30 to 6.87) meters and the
to Arduino hardware. Nearly all Arduino modules are
average error is (2.69) meters. While in speed of (100) km, the
compatible with this open-source tool, that is simple to install
error range changes between (30.56 to 31.89) meters and the
and utilize to start creating code while on the go. This
average error is (31.39) meters.
environment supports C and C++. It runs on the Java Platform,
that provides built-in functions and commands that are Table II. SHOWS ERROR IN LOCATION DETERMINATION OF THE SECOND
necessary for debugging, editing, and compiling the code in POINT FOR THE PROPOSED SYSTEM
the environment, and is easily available for operating systems
including MAC, Windows, and Linux [13][14]. Coordinates from Google Lat Long
Maps for the correct location 36.3696310000 43.1286010000
b) TinyGPS Arduino library Error/
Speed Error
Lat. Long. Lat. Long. m
km /m
It offers GPS information from the National Marine avg.
Electronics Association, including position, altitude, time, 36.3696174 43.1286201 2.3
date, etc. To save resource use, the library ignores all but a few 0 36.3696632 43.1286048 3.5 36.36965560 43.12859725 2.69
36.3696861 43.1285667 6.8
essential GPS variables and stays away from floating point 36.3696823 43.1287384 13.5
[15]. 40 36.3697090 43.1287193 13.6 36.36970138 43.12874221 14.83
36.3697128 43.1287765 18.0
V. RESULTS OF THE PROPOSED GPS AND DISCUSSION 36.3697090 43.1288528 24.0
After connection, the system is turned on and after many 60 36.3697204 43.1288375 23.3 36.36972045 43.12884521 23.98
36.3697357 43.1288490 25.0
seconds the antennas of the sensors will take the signals from
36.3697013 43.1289062 28.3
the satellites. 80 36.3697814 43.1288948 31.0 36.36974716 43.12889099 28.91
36.3697471 43.1288719 27.4
The major objective of the project was to gather field data
36.3697357 43.1289176 30.5
in real-world scenarios for GPS users operating land vehicles. 100 36.3697395 43.1289215 31.0 36.36973953 43.12892532 31.39
In the city of Mosul/ Iraq, there are 1.683000 people living in 36.3697433 43.12892913 31.8
it, which has an area of 180 km2. It is a region with a Thus, the importance of using more than one GPS sensor
diversified economy, several businesses, agricultural areas, a is obvious that is to reduce the effect of multipath on location
number of commercial structures, and a significant number of determination and improve the accuracy of the GPS system,
suburban homes, making it an appropriate location for testing
since the average error rate is approximately (21.99 and 31.39)
the proposed system.
meters at 100 Km speed for the considered two points
Two points with coordinates (36.3807700000, respectively. These errors are a reasonable due to the weather
43.1655780000) and (36.3696310000, 43.1286010000) in the conditions, and the problems of transmitting the signal from
city are taken to apply the proposed system on them, Table I the satellite to the GPS sensor, and also the use of hardware is
shows the errors in the first point coordinates and for (5) somewhat cheap.
speeds on the same point. In speed of (0) km, the error range
Figure 6 (a and b) shows the results of the proposed system
changes between (7.70 to 11.84) meters and the average error
on Google earth, where (a) represents one GPS sensor and (b)
is (9.89) meters. While in speed of (100) km, the error range
three GPS sensors. In one sensor state, it is possible to see the
changes between (21.70 to 22.32) meters and the average error
amount of error in the map, where the points are very far apart
is (21.99) meters.
from each other, but in the case of using three GPS sensors, it
Table I. SHOWS ERROR IN LOCATION DETERMINATION OF THE FIRST POINT can be seen that the points are close. This indicates that the
FOR THE PROPOSED SYSTEM multipath has been reduced to an acceptable, reasonable and
developmental extent.
Coordinates from Google Maps Lat Long
for the correct location 36.3807700000 43.1655780000
Speed Lat. Long. Error Lat avg. Long avg. Error/
km /m m avg.
36.38067245 43.16563034 11.84
0 36.38068771 43.16561889 9.89 36.38068771 43.16561889 9.89
36.38070678 43.16561126 7.70
36.38071441 43.16571044 13.35
40 36.38072967 43.16567993 10.12 36.38071441 43.16569519 12.18
36.38069915 43.16569519 13.11
36.38065338 43.16553878 13.49
60 36.38063812 43.16553115 15.26 36.38064956 43.16553115 14.09
36.38066482 43.16554641 12.12
36.38072586 43.16579055 19.62
80 36.38078689 43.16577529 17.72 36.38075256 43.16578674 18.72
36.38074874 43.16579055 19.13
36.38073730 43.16581726 21.70 (a)

32
International Conference on Engineering, Science and Advanced Technology (ICESAT-2023), Technical Engineering
College / Mosul, Iraq

[6] E. Yoder and E. Humphreys, “Low-Cost Inertial Aiding for Deep-

Urban Tightly-Coupled Multi-Antenna Precise GNSS”,
arXiv:2201.11776v1 [eess.SP], pp. 1-16, Jan 2022.
[7] E. Kaplan and J. Hegarty, Understanding GPS: principles and
applications, Artech House, 2005.
[8] K. Borre, M. Akos, N. Bertelsen, P. Rinder and H. Jensen, A software-
defined GPS and Galileo receiver: a single-frequency approach,
Springer Science & Business Media.
[9] T. Luo, Y. Gao, H. Wang and Z. Liu, The impact of multipath on GPS
and its mitigation techniques. Measurement, Vol. 131, pp. 636-647,
2019.
[10] V. Ta and Y. Lee, “Advanced signal processing techniques for
mitigating multipath effects in GNSS”, IEEE Communications
Magazine, Vol. 54, No 3, pp. 82-89, 2016.
(b) [11] K. Kaswan, S. Singh and S. Sagar, “Role Of Arduino In Real World
Applications”, Volume 9, Issue 01, pp. 113-1116, January 2020.
Fig. 6. Google earth results (a) for one sensor GPS (b) for the proposed three [12] Y. Zhao, Y. Zhan and H. Zhou, “Design of GPS Information Processing
sensors GPS System Based on Single Chip Microcomputer”, International Journal
of Emerging Trends in Social Sciences, Vol. 5, No. 1, pp. 12-15, 2019.
In comparison with the omnidirectional case, the [13] M. Fezari and A. Al Dahoud, “Integrated Development Environment
directivity (or gain) is increased (approximately doubled) “IDE” For Arduino”, unpublished.
when three antennas are used. [14] D. Abdulahad Aziz, S. Saadoon, S. Alani and S. Nozad, “Design and
Implementation of GPS Based Quadcopter Control System”, IMDC-
VI. CONCLUSIONS SDSP 2020, June 28-30, 2020.
[15] X. Hong, W. Wang, N. Chang and Q. Yin, “A subspace based code
Because of the significance of GPS usage, several tracking loop design for GPS multi antenna receiver in multipath
strategies have been created throughout time to address this environment”, GPS Solutions, Vol. 24, No. 109, pp. 1-17, December,
issue. In this paper, a simple and low-cost tracking device; 2019.
consists of three Neo-6m GPS sensors With Arduino Mega
microcontroller is designed and implemented to reduce the
effect of multipath problem. With the help of three GPS
sensors, the receiver takes the average coordinates for them in
a short time. Thus, the system performance was tested for two
locations, in terms of error distance, while the system is
stationary (zero speed) in addition to four other speeds. It has
been proven that system performance does indeed depend on
locations with obstacles such as buildings which will reduce
system accuracy and the GPS receiver movement and the
number of GPS satellite signals that the GPS receiver can
receive. Moreover, it has been shown that faster driving
motion produces a relatively higher average error distance
compared to slower walking motion. For the proposed multi-
antenna GPS and at speed of 100 km, the average error in the
location of point one and point two are 21.99 and 31.34 meters
respectively. In comparison with single-antenna GPS, the
percentage improvement in the location accuracy is 1.50%
and 1.59% for the two points respectively.
REFERENCES
[1] M. Kasliwal and H. Patil, “Smart location tracking system for dementia
patients”, International Conference on Advances in Computing,
Communication and Control (ICAC3), pp 1–6, 2017.
[2] R. Sun, G. Wang, W. Zhang, L. Hsu and W. Ochieng, “A gradient
boosting decision tree based GPS signal reception classification
algorithm”, Applied Soft Computing Journal, in press.
[3] V. Carvalho and A. Moraes, “Modeling multifrequency GPS
multipath fading in land vehicle environments”, GPS Solutions, Vol.
25, No. 3, pp. 1-14, October 2019.
[4] E. Amani, K. Djouani, A. Kurien, J. De Boer, W. Vigneau and L. Ries,
“GPS Multipath Detection in the Frequency Domain”, unpublished.
[5] X. Hong, T. Gan, M. Lin, N. Chang, W. Wang and Q. Yin, “A
Subspace-based Spatial and Temporal Multipath Mitigation for Multi-
antenna GPS Receiver”, University of Exeter, pp. 1256-1266, June,
2020.

33