=Paper=
{{Paper
|id=Vol-3097/paper18
|storemode=property
|title=RSS-Based Fusion of UWB and WiFi-Based Ranging for Indoor Positioning
|pdfUrl=https://ceur-ws.org/Vol-3097/paper18.pdf
|volume=Vol-3097
|authors=Ghazaleh Kia,Jukka Talvitie,Laura Ruotsalainen
|dblpUrl=https://dblp.org/rec/conf/ipin/KiaTR21
}}
==RSS-Based Fusion of UWB and WiFi-Based Ranging for Indoor Positioning==
https://ceur-ws.org/Vol-3097/paper18.pdf
RSS-Based Fusion of UWB and WiFi-Based Ranging
for Indoor Positioning
Ghazaleh Kia1 , Jukka Talvitie2 and Laura Ruotsalainen3
1
University of Helsinki, Department of Computer Science, Helsinki, Finland
2
Tampere University, Unit of Electrical Engineering, Tampere, Finland
3
University of Helsinki, Department of Computer Science, Helsinki, Finland
Abstract
WiFi positioning with estimated ranges using Round Trip Time (RTT) measurements based on IEEE
802.11 Wireless Local Area Network (WLAN) has become well known since Fine Timing Measurement
(FTM) protocol has been characterized. However, the multipath effect is one of the barriers to accurate
time-based range measurement. On the other hand, Ultra Wide Band (UWB)-based range measurement
has fair resistance to multipath effects but its accuracy is highly dependant on the orientation of the
antennas in the transmitter and the receiver and its transmit power is also limited due to the applied
regulations. This paper utilizes a Received Signal Strength (RSS)-based fusion of both UWB and WiFi-
based range measurements to increase the indoor positioning accuracy. The proposed method takes the
advantage of WiFi FTM protocol as well as Two-Way Ranging (TWR) for UWB devices. The empirical
range measurement campaign is done at the University of Helsinki premises. Test points with known
positions are considered as the ground truth to evaluate the results. The outcome proves that fusing
UWB and WiFi ranges for indoor positioning, improves the accuracy in comparison with using the UWB
or WiFi alone.
Keywords
Indoor Position Estimation, Received Signal Strength (RSS), Trilateration, UWB Two-Way Ranging
(TWR), Wi-Fi Fine Timing Measurements (FTM), Wi-Fi Round-Trip Time (RTT)
1. Introduction
Signals-of-opportunity (SOOP) provide a good means for positioning indoors, where Global
Navigation Satellite System (GNSS) signals are not available [1]. One of the well-known SOOP
signals used for indoor positioning is WiFi [2]. There are two commonly-used methods to utilize
WiFi signals for indoor positioning including RSS-based and Time-of-Flight (ToF)-based methods.
The most conventional RSS-based procedure for WiFi indoor positioning is fingerprinting [3].
This method includes two phases: offline and online. In the offline phase, RSS values received
from fixed anchors at each known location are collected in a data set, mapping the locations to
the collected RSS values. Then, in the online phase, the online collected RSS values are compared
with the previously collected RSS values in the data set and the corresponding position or an
average of positions is reported as the position of the user [4]. However, acquiring the data set
IPIN 2021 WiP Proceedings, November 29 β December 2, 2021, Lloret de Mar, Spain
" ghazaleh.kia@helsinki.fi (G. Kia); jukka.talvitie@tuni.fi (J. Talvitie); laura.ruotsalainen@helsinki.fi
(L. Ruotsalainen)
0000-0001-7010-986X (G. Kia); 0000-0001-7685-7666 (J. Talvitie); 0000-0002-4057-4143 (L. Ruotsalainen)
Β© 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073 CEUR Workshop Proceedings (CEUR-WS.org)
is time-consuming and expensive. As another approach, WiFi can be deployed for positioning
purposes by getting the RSS values to find the range between the transmitter and the receiver
by signal propagation models. This method is vulnerable to variations in range measurement
due to the propagation model errors [5]. To make localization easier using WiFi, IEEE released
the new version of the 802.11 WLAN standard with a protocol of FTM in 2016. This protocol
enables WLAN devices to find the distance between each other using RTT measurements [6].
RTT is the time it gets for a signal to travel from a user to an access point and travels back
to the user after a fixed interval. RTT does not require any clock synchronization since the
access point transmission is triggered by the signal of the user and the user clock offset cancels
between the incoming and outgoing transmissions [4]. However, RTT is not free of problems.
It provides a time-delay-based range and finding an accurate delay-based range in indoor
environments with dense-multipath propagation is challenging [5]. To address the multipath
effect, a general Bayesian Filter was presented in [7] to integrate RTT range measurement with
map information. By this technique, authors were able to filter out the RTT-based range errors
and provide a position solution with an accuracy of 3 π. In 2019, Dvorecki et al. in [5] have
used a supervised deep learning method to estimate the RTT-based range in a dense-multipath
environment. For this purpose, the authors find the first observed LOS signal path by Channel
Impulse Response (CIR). Then, by having the channel estimation vector, they estimate the
delay of the first observed path by Siamese Artificial Neural Network (ANN). The results prove
that the ranging accuracy was increased to 4 π using ANN. Another technique presented in
2019 was the use of a hybrid WiFi RTT and WiFi RSS [8]. The authors measured the ranges
based on both RTT and RSS and then fused the ranges using Kalman Filter to get an accurate
position solution. To evaluate their work, authors have considered a set of test points with
known locations in an indoor area of 192 π2 . The results show that the position accuracy
achieved in an indoor environment was 1.435 π. In 2020, the creators of WiNar have presented
a combination of both RTT-based fingerprinting and ranging methods to mitigate the effects of
multipath on time-based ranging. The results show that this work outperforms using RTT-based
ranging or RSS-based fingerprinting alone [9]. The authors have reached an accuracy of 86 ππ.
However, in WiNar the RTT-based fingerprints should be collected and the environment-specific
database should be created, which is expensive and time-consuming. Retscher et al. in [10]
have presented the idea of using UWB nodes to fix the errors in WiFi RSS-based positioning.
In this method, the indoor building environment is divided into different parts, and passing
any UWB base stations shows that the user is in which part of the whole building. Then the
results are integrated with the ranges calculated using RSS values from visible WiFi base stations
and theoretical path loss models. Using this method, the authors were able to achieve 2.65 π
accuracy.
UWB-based positioning is a rapid developing field regarding the accuracy and the robust-
ness it provides [11]. The accuracy in both static and dynamic positioning on reference test
points are investigated in [12] and [13]. The static position investigation is done to analyze
the ranging capabilities in a favorable environment. Static positioning can also be used in
Wireless Sensor Network (WSN) nodes positioning, where the sensor nodes are fixed [14].
The range measurement in static condition using Pozyx devices is also addressed in [15]. The
results indicate that the UWB-based range measurement can have errors as big as 1.73 π and
positioning can have errors more than one meter in a dense indoor environment where many
objects are causing multipath effects. This error can be increased to 5.919 π when the user
is dynamic in a narrow corridor [13]. Furthermore, the accuracy of UWB devices where the
antenna is omnidirectional is highly dependent on the relative orientation of the transmitter
and the receiver [1]. In this work, we are utilizing Pozyx developer anchors and tag, which
utilize omnidirectionally radiating monopole UWB antennas [16]. However, in real positioning
scenarios, it is highly probable that at some points the transmitter and the receiver directions
are not heading toward each other. This can result in poor range measurement and position
estimation with errors [1]. Furthermore, the ranging error of WiFi RTT tested in static positions
could be as high as 1.494 π even after bias removal [17]. Similarly in [18] static test points are
used to evaluate the RTT-based ranging calibration. The results show that the position error in
test points could be as high as 1.55 π. However, if the user is dynamic, using WiFi RTT for
positioning is shown to have an error as high as 4 π [5].
In this work, considering the UWB and WiFi RTT-based positioning, we focus on mitigating
the effects of the above-discussed issues, including the limitation of transmission power and
orientation-dependent accuracy of the UWB TWR, as well as multipath and position-dependent
errors with the WiFi RTT. We address the error in the WiFi RTT device, present an RSS-based
fusion of the UWB TWR and the WiFi RTT, and evaluate our work by using multilateration for
static positions [8, 12, 13, 17, 18].
The remainder of the article is divided into the following sections. Section 2 explains the
theoretical background for range measurements and multilateration. The methodology of
deploying the range measurement, correcting WiFi position-dependent error, and RSS-based
fusion method are presented in Section 3. Section 4 provides the experiments and results,
followed by the Section 5 concluding this work.
2. Background
Radio Frequency (RF) signals are widely used for indoor positioning purposes. In this work, we
address trilateration using radio-based range measurements from three anchors. Regarding the
indoor environment, we investigate the two candidate signals [10]: UWB and WiFi.
2.1. WiFi RTT using FTM Protocol
Regarding the rapid use of WiFi signals in indoor positioning, FTM protocol enables an initiating
device like a smartphone to find its distance to a WiFi base station like an Access Point (AP).
The process begins when an initiator (it can be a mobile phone) sends an FTM request to an AP.
If the AP supports the FTM process and can respond, it is called a Responder. The Responder
may agree or refuse the initiatorβs message for the ranging procedure. If it agrees with the
message, it will send an acknowledgment message (ACK) to the initiator. Then the Responder
starts sending the initiator an FTM message and waits for the ACK from the initiator. The AP
is allowed to send the next message only when it receives the ACK from the initiator. This
procedure is illustrated in Fig. 1.
Considering the transmission timestamp of the FTM message and the reception of the ACK
in the ππ‘β procedure of a burst, RTT is estimated:
Initiator Responder
(Smartphone) (AP)
π‘1,1
π‘2,1
FTM
π‘3,1 π‘4,1 Interchange 1
π‘2,2 π‘1,2
FTM
π‘3,2 Interchange 2
π‘4,2
π‘1,3
π‘2,3
FTM
π‘3,3 Interchange 3
π‘4,3
Figure 1: An overview of WiFi FTM protocol: one burst with 3 FTM exchange
ππ
π π = (π‘4,π β π‘1,π ) β (π‘3,π β π‘2,π ) (1)
where, π‘4,π is the receiving time of ACK, π‘1,π is the transmission time of the FTM message, and
π‘3,π β π‘2,π is the processing time in the mobile phone, as illustrated in Fig. 1. Several FTM-ACK
processes occur in one burst. The RTT calculation continues for all the FTM-ACK processes in
the burst and finally, to report the RTT, the average value will be considered.
All the timestamps are transmitted to the mobile phone to calculate the RTT. In this way, the
user is enabled to estimate its position based on RTT-based ranges while its privacy is preserved
[9]. The RTT shown in (1) is used to calculate the range as follows:
1
π
= Γ ππ
π π Γ π (2)
2
where π is the speed of the light. The user measures the RTT to all the responders in range,
which enables the user to multilaterate its position [8]. In this work, we use trilateration to find
the user position.
2.2. Trilateration
The trilateration method uses triangles geometry to estimate the target position of the mobile
objects. For estimating the 2D position using multilateration, at least three base stations/anchors
are required [8]. As illustrated in Fig. 2, the position of a user equipment π = (π₯, π¦) can be
estimated from (3) as:
(x1,y1) (x3,y3)
R2
(x2,y2)
Base station with known location
User with unknown location
Figure 2: Positioning using anchors with known locations
P = (Kπ K)β1 Kπ J (3)
where β‘ β€
π₯1 β π₯2 π¦1 β π¦2
.. .. β₯ (4)
K = 2β£ . . β¦
β’
π₯ 1 β π₯ π π¦1 β π¦π
and
π
22 β π
12 β (π₯22 β π₯21 ) β (π¦22 β π¦12 )
β‘ β€
.. (5)
J=β£ .
β’ β₯
β¦
π
π2 β π
12 β (π₯2π β π₯21 ) β (π¦π2 β π¦12 )
Moreover, (π₯1 , π¦1 ), ..., (π₯π , π¦π ) are the known positions of the π anchors, and π
1 , ..., π
π are
the range measurements from the user to the π anchors.
2.3. Range Measurement using UWB TWR
UWB is a microwave signal comprising electromagnetic radiation within a frequency range
from 3.1 πΊπ»π§ to 10.6 πΊπ»π§. UWB signals are shown to propagate through walls and different
materials [19]. Range measurement using UWB can be done with different techniques relying
on time measurements such as Time of Arrival (ToA), Time Difference of Arrival (TDoA), and
TWR. In this paper, the considered TWR approach depends on the time that RF signal requires
to travel from the transmitter to the receiver, to be processed, and travel back to the transmitter.
In this way, synchronization issues are conveniently solved in comparison with ToA and TDoA
method, [4]. In TDoA method, the anchors are required to be accurately synchronized since the
position estimation is done by finding the difference between the time stamps of the arriving
signal at the anchors [20]. By using the TWR method, the need for anchors synchronization is
eliminated.
Time-based measurement can be divided into three principles including continuous-wave,
impulse radio, and pseudo-noise modulation. The most commonly used method is the impulse
radio. In this method, to have accurate timing a large bandwidth is required to have a narrow
pulse since wide pulse results in capturing all the reflections from the environment while the
signal is scattered onto the objects such as walls [12]. Regarding the fact that in UWB systems
the pulse width is in order of nanosecond since the bandwidth is around 500 π π»π§, UWB
can filter reflections of the signals to a great extent. However, the interference potential with
conventional radio systems should be avoided by limiting the power spectral density. Therefore,
regarding the decision of the Federal Communications Commission (FCC) in the USA and the
Harmonized European Standard (EU ETSI EN 302 065), the maximum allowed isotropic radiated
power density for unlicensed use of UWB is limited to β41.3 ππ΅π/π π»π§.
In the next section, we describe how we utilize the smartphone and the Pozyx developer tag
to collect the range measurements.
3. Methodology
In this paper, we are using both UWB and WiFi anchors for obtaining the desired position
estimates. To reach our goal, we have put a pair of UWB and WiFi anchors aligned on each
other in three known locations, and the ranges from the user equipment to the anchors are
calculated with both UWB and WiFi devices separately. The instruction of setting up the system
is illustrated in Fig. 3 for the user equipment and one of the anchors.
UWB (Pozyx Tag) UWB-based Range measurement
UWB
Connected (saving UWB TWR
to a laptop (Pozyx Anchor)
on laptop)
User Equipment WiFi-based Range measurement Anchor
Google Smartphone
(Unknown Location) WiFi WILD (Known Location)
(collecting WiFi RTT)
Tag and Smartphone UWB and WiFi Anchors
Aligned Aligned
To make the WiFi-based and
UWB-based positions comparable
Figure 3: Instruction on setting up the system
The range measurements are collected in the reference points and averaging is used for the
collected data at each point.
3.1. UWB Range Collection
To collect UWB range measurements, Pozyx developer tags are used. These tags have a De-
cawave DW1000 transceiver, which can operate within a frequency range from 3.5 πΊπ»π§ to
6.5 πΊπ»π§, and the power gain can be modified up to a maximum of 33 dB to control the output
power. However, to assure that the signal transmitted power density stays below the maximum
allowed power, the Pozyx tags average transmit power gain is set below 20 ππ΅. The ranging
method used for finding the distances between the Pozyx tag and the anchor is TWR. The Pozyx
tag at the user equipment is connected to a laptop to save the range measurements on the laptop
and get the timestamp from the operating system.
3.2. WiFi Range Collection
To collect the WiFi RTT measurements, we used a Google Pixel 3 smartphone to receive the
data from WLAN anchors. This phone can support the Android RTT API as the receiver to
collect the range measurements [17, 18]. The WiFi ranges and the timestamps are collected and
saved in the smartphone.
3.3. WiFi Position-Dependent Error in Collected Ranges
By investigating the WiFi range measurements, we recognized that there is a non-constant
error value for the measured ranges. Based on [21], there are two equipment-related issues with
the WiFi devices. The first one is an offset value in the reported range measurements, which is
quite repeatable and can be estimated by taking many measurements in known distances and
calculating the average value. The second one is the position-dependent error in the measured
ranges. To solve the latter, several measurements of the range should be done in an environment
without any obstruction and find a linear fit to model the error [22]. Thus, we have done a
range-measurement test inside the Anechoic Chamber of the Department of Computer Science
at the University of Helsinki. The Anechoic chamber is a room that absorbs electromagnetic
reflections and is isolated from interfering signals [23]. The Anechoic chamber of the department
has a size of 250 ππ Γ 240 ππ Γ 400 ππ and is constructed using the Rainford EMC Systems
modular panel system. The whole chamber, including the sidewalls and the ceiling, is covered
with 30.5 ππ thick, solid, sharp-tip, resistive pyramidal foam absorbers. The chamber is shown
in Fig. 4.
To do the experiment and analyze the offset and position-dependent errors, we have fixed
one of the WiFi devices in the chamber and put the Google Android Phone in different distances
from 0 to 200 ππ with 10 ππ increasing steps. We collected the range measurements with the
Android phone for three minutes at different distances. Then we averaged the collected range
measurement in each known distance to model the distances. Least-squares line fitting is used
to estimate the calibrated distances based on the collected RTT-based ranges. The linear model
is shown in Fig. 5, where the fitted first-order polynomial is defined as:
π
ππππππππ‘ππ = 0.614 Γ π
πππππππ‘ππ + 4.266 (6)
Using this model, we were able to estimate the correct ranges with a mean error of 35 ππ.
Figure 4: Collecting WiFi RTT measurements inside the Anechoic Chamber
-3.5
Average Reported Distance Linear Fit
-4
Average Reported Distance (m)
-4.5
-5
-5.5
-6
-6.5
-7
0 0.5 1 1.5 2
Real Distance (m)
Figure 5: Use of linear fit for modeling the position-dependent error of the WiFi RTT-based range
measurements
3.4. UWB and WiFi Fusion
To fuse the WiFi and UWB range measurements, we have first synchronized the smartphone
and the laptop (to which the Pozyx tag is connected) by Network Time Protocol (NTP). NTP
is commonly used to synchronize the clock of network devices within the accuracy of a few
milliseconds. By synchronizing the smartphone, which collects the WiFi RTT, with the laptop,
which collects the UWB TWR measurements, we can fuse the measurements occurring with
the same timestamp.
We take the advantage of corresponding RSS values for each range measurement. Regarding
the fact that path loss and attenuation result in decreased RSS values [24], and knowing that
very high RSS, in many environments indicates LOS dominated channel[5], we have defined the
weights for each range measurement and calculated the weighted average of the two signals.
The normalized RSS values are the parameters for defining the weights. The following equations
are used to calculate the weights for fusing the measurements.
π
πππ π π΅ π2 Γ ππ ππΉ π
π1 = ( βοΈπ’π ) Γ π2 π π΅ 2 (7)
π=1 π
ππ(π) ππ π΅ π π π π΅ + ππ ππΉ π
π
πππ ππΉ π π 2 ππΉ π Γ ππ π π΅
π2 = ( βοΈπ’π )Γ π (8)
π=1 π
ππ(π)π ππΉ π ππ2 π π΅ + ππ2
ππΉ π
where π
πππ π π΅ and π
πππ ππΉ πβοΈ are the corresponding RSSβοΈπ’πvalues for each range measurement
π’π
in π€ππ‘π‘π , π’π is the update rate, π=1 π
ππ(π)π π π΅ and π=1 π
ππ(π)π ππΉ π are the sum of RSS
values at each time epoch for UWB and WiFi, respectively.
In the definition of π1 and π2 the effect of RSS and accuracy are applied. In one time epoch,
the number of available measurements equals the update rate of the device π’π. Considering the
processing of data after each time epoch while we have all the measurements occurring in one
epoch, we can define the fraction βοΈπ’ππ
ππ for each measurement. Larger RSS values imply
π=1 π
ππ(π)
a stronger signal and result in larger π1 and π2 values. In the second fraction of the equations
(7) and (8), the effect of accuracy is applied. According to ISO 5725-1, accuracy is defined by
two factors: trueness and precision, where trueness is the mean and precision is the variance of
the error distribution. For applying the two factors in the RSS-based fusion, we have first found
the mean and variance of the estimated position errors using UWB (ππ π π΅ , ππ π π΅ ), as well as
those of the estimated position errors using WiFi (ππ ππΉ π , ππ ππΉ π ). Then we have defined the
weights based on the mean of a Gaussian Probability Distribution Function (PDF) which is the
product of the two (WiFi and UWB) Gaussian PDFs.
Finally, the weights will be calculated using (9) and (10).
π1
ππ π π΅ = (9)
π1 + π2
π2
ππ ππΉ π = (10)
π1 + π2
Thus, the final fused range value would be:
πΉ π’π πππ
ππππ = ππ ππΉ π Γ π
π ππ π + ππ π π΅ Γ π
π π π΅ (11)
where π
π ππΉ π represents the collected range measurements with WiFi devices, which are
calibrated by using (6), and π
π π π΅ represents the collected range with the UWB devices.
4. Experiments and Results
To test the algorithm and evaluate our work, a measurement campaign is done at the premises
of the Department of Computer Science, University of Helsinki. The floor plan of the area for
the test is illustrated in Fig. 6.
8m
Table Table
10 m
Anchors
Reference Points
Figure 6: Floor Plan of the area of data collection
18 reference points with known locations are considered. The reference points are marked
with yellow signs as illustrated in Fig. 7 and 8.
Figure 7: Reference points for evaluating the results inside the hall.
The WiFi anchors are Compulab WILD WLAN devices that are capable of providing RTT
measurements. The UWB anchors are Pozyx Developer anchors. The aligned UWB and WiFi
anchors in one of the base stations locations are illustrated in Fig. 9.
For the UWB modules, channel 5 is selected, preamble length is 1024, Pulse Repetition
Figure 8: Reference points for evaluating the results inside the hall and the corridor.
Figure 9: One of the base stations including both WiFi and UWB anchors
Frequency (PRF) is 64 π π»π§, bitrate is 110 ππππ , and transmit power is 19.5 ππ΅.
The Pozyx tag is connected to a laptop for UWB TWR collection using a developed Python
code. Furthermore, the smartphone using a developed Android application is utilized for WiFi
RTT collection. The phone and the tag are illustrated in Fig. 10. The range data is collected
once using the synchronized smartphone and the laptop. The UWB and WiFi measurements
are used both separately and fused for estimating the positions.
Figure 10: Pozyx tag and smart phone for data collection
4.1. WiFi-RTT based ranges for estimating the positions
Using the range measurements collected at the test points, we have first done the position
estimation using WiFi-based ranges and the trilateration method. The estimated positions are
illustrated in Fig. 11.
The calculated mean error of the estimated positions using only the WLAN devices is
133.9 ππ.
4.2. UWB-TWR based ranges for estimating the positions
After using the WiFi-based ranges, we have used trilateration for UWB-based range measure-
ments. The estimated positions using only UWB TWR are illustrated in Fig. 12.
Estimating the positions using only the UWB devices resulted in a mean error of 111.4 ππ.
4.3. RSS-based fused range measurements for estimating the positions
After investigating the UWB and WiFi-based position estimations, we have fused the range
measurements using the RSS-based algorithm presented in the methodology section. By using
the fused range measurements, we were able to decrease the mean error to 75.7 ππ. The
estimated positions using the fused UWB and WiFi-based range measurements are illustrated
in Fig. 13.
The results prove that UWB and WiFi fusion presents sub-meter level accuracy in a dense
multipath indoor environment where the WiFi signals are experiencing reflections as well as
position-dependent errors, and the omnidirectional antennas of Pozyx UWB devices are not
1000
Estimated Positions Reference Points Anchors
800
600 5 6
5 4 7
6
400 47 3 8
y(cm)
38 2 9
9 2
200 11 10
11 10
11 17
12
1415 18
0 1312 16
13 14 15 16 17 18
-200
-400
-100 0 100 200 300 400 500 600 700
x(cm)
Figure 11: Estimated positions using WiFi-based range measurements
1000
Estimated Positions Reference Points Anchors
800
600 55 66
4 7
74
400 3 3 8
y(cm)
8
2 29 9
200 1 1 10
10
11
0 12
12
11
13
13 14 15
15 16 1717 18
14 16 18
-200
-400
-400 -200 0 200 400 600 800
x(cm)
Figure 12: Estimated positions using UWB-based range measurements
necessarily heading toward each other. As summarized in Fig. 14, compared to individual UWB
and WiFi-based measurements, the mean and Standard Deviation (STD) of the positioning
error are decreased by 32 β 43% and 39 β 51%, respectively, by fusing the UWB and WiFi
measurements.
1000
Estimated Positions Reference Points Anchors
800
600 5 6
5
64 7
4 7
400 38 3 8
229 9
200 11 10
10
11
12
0 12 17
13 14 15
13 14 1516 16 18 17 18
-200
-400
-100 0 100 200 300 400 500 600 700
x(cm)
Figure 13: Estimated positions using RSS-based fused UWB and WiFi-based range measurements
Errors in Position Estimation (cm)
160
140
120
100
80
60
40
20
0
Mean STD Mean STD Mean STD
UWB TWR WiFi RTT Fused-Range
Figure 14: Errors in estimated positions using different measurements
5. Conclusion
In this work, the WiFi RTT-based ranging and UWB TWR are addressed using Compulab
WLAN and Pozyx UWB devices. The range measurements are first used in the multilateration
algorithm to estimate the position solution separately for the WiFi and UWB. Regarding the
problem of multipath effect which is highly effective on WiFi RTT-based ranges and the problem
of orientation-dependent accuracy of UWB as well as transmit power limitations by regulations,
an RSS-based fusion of the range measurements is used to investigate the superiority of the
integrated ranges for the position estimation. The results confirm that the accuracy of estimated
static position solutions is improved in the real indoor environment by fusing the UWB and
WiFi-based range measurements. In the next steps of this research, specific noise probability
distribution in the range measurements will be addressed. In addition, a mobile user with
appropriate tracking algorithms is considered for a navigation application. Besides the above, it
is important to investigate a larger positioning area with additional reference points for more
comprehensive testing of the system behavior.
Acknowledgments
This work was funded by the European Space Agency ESA AO/2-1716/19/NL/CRS/hh and the
University of Helsinki.
References
[1] P. Figueiredo E Silva, Signals of opportunity for positioning purposes, Ph.D. thesis, Tampere
University of Technology, 2018. URL: http://urn.fi/URN:ISBN:978-952-15-4280-0.
[2] C. Gentner, M. Ulmschneider, I. Kuehner, A. Dammann, WiFi-RTT indoor positioning,
2020, pp. 1029β1035. doi:10.1109/PLANS46316.2020.9110232.
[3] G. Mendoza-Silva, A. C. Costa, J. Torres-Sospedra, M. Painho, J. Huerta, Environment-
aware regression for indoor localization based on WiFi fingerprinting, IEEE Sensors
Journal (2021) 1β1. doi:10.1109/JSEN.2021.3073878.
[4] P. Groves, Principles of GNSS, inertial, and multisensor integrated navigation systems,
Second edition, 2013.
[5] N. Dvorecki, O. Bar-Shalom, L. Banin, Y. Amizur, A machine learning approach for Wi-Fi
RTT ranging, 2019. doi:10.33012/2019.16702.
[6] (Revision of IEEE Std 802.11-2012) - Part 11: Wireless LAN Medium Access Control (MAC)
and Physical Layer (PHY) specifications, December 7th, 2016.
[7] L. Banin, U. Schatzberg, Y. Amizur, WiFi FTM and Map information fusion for accurate po-
sitioning, in: 2016 International Conference on Indoor Positioning and Indoor Navigation
(IPIN), 2016.
[8] G. Guo, R. Chen, F. Ye, X. Peng, Z. Liu, Y. Pan, Indoor Smartphone Localization: A Hybrid
WiFi RTT-RSS Ranging Approach, IEEE Access 7 (2019) 176767β176781. doi:10.1109/
ACCESS.2019.2957753.
[9] O. Hashem, M. Youssef, K. A. Harras, WiNar: RTT-based Sub-meter Indoor Localization
using Commercial Devices, in: 2020 IEEE International Conference on Pervasive Comput-
ing and Communications (PerCom), 2020, pp. 1β10. doi:10.1109/PerCom45495.2020.
9127363.
[10] G. Retscher, V. Gikas, H. Hofer, H. Perakis, A. Kealy, Range validation of UWB and
Wi-Fi for integrated indoor positioning, Applied Geomatics 11 (2019). doi:10.1007/
s12518-018-00252-5.
[11] M. MΓ€kelΓ€, M. Kirkko-Jaakkola, J. Rantanen, L. Ruotsalainen, Proof of concept tests on coop-
erative tactical pedestrian indoor navigation, in: 2018 21st International Conference on In-
formation Fusion (FUSION), 2018, pp. 1369β1376. doi:10.23919/ICIF.2018.8455380.
[12] P. Dabove, V. Di Pietra, M. Piras, A. A. Jabbar, S. A. Kazim, Indoor positioning using
Ultra-wide band (UWB) technologies: Positioning accuracies and sensorsβ performances,
in: 2018 IEEE/ION Position, Location and Navigation Symposium (PLANS), 2018. doi:10.
1109/PLANS.2018.8373379.
[13] V. D. Pietra, P. Dabove, M. Piras, A. Lingua, Evaluation of positioning and ranging errors
for UWB indoor applications, in: IPIN, 2019.
[14] J. Grigulo, L. B. Becker, Experimenting sensor nodes localization in WSN with UAV acting
as mobile agent, in: 2018 IEEE 23rd International Conference on Emerging Technologies
and Factory Automation (ETFA), volume 1, 2018, pp. 808β815. doi:10.1109/ETFA.2018.
8502536.
[15] V. Di Pietra, P. Dabove, M. Piras, Loosely coupled GNSS and UWB with INS integration
for indoor/outdoor pedestrian navigation, Sensors 20 (2020). doi:10.3390/s20216292.
[16] Pozyx knowledge center-Hardware setup, https://docs.pozyx.io/creator/latest/
getting-started/hardware-setup/, 2021. [Online; accessed 5-May-2021].
[17] C. MA, B. Wu, S. Poslad, D. R. Selviah, Wi-Fi RTT ranging performance characterization
and positioning system design, IEEE Transactions on Mobile Computing (2020) 1β1.
doi:10.1109/TMC.2020.3012563.
[18] H. Cao, Y. Wang, J. Bi, S. Xu, M. Si, H. Qi, Indoor positioning method using WiFi RTT
based on LOS identification and range calibration, ISPRS International Journal of Geo-
Information 9 (2020). doi:10.3390/ijgi9110627.
[19] M. MΓ€kelΓ€, J. Rantanen, J. Ilinca, M. Kirkko-Jaakkola, S. Kaasalainen, L. Ruotsalainen, Co-
operative environment recognition utilizing uwb waveforms and cnns, in: 2020 European
Navigation Conference (ENC), 2020, pp. 1β8. doi:10.23919/ENC48637.2020.9317403.
[20] W. Wang, J. Huang, S. Cai, J. Yang, Design and implementation of synchronization-free
tdoa localization system based on uwb, Radioengineering 27 (2019) 320β330. doi:10.
13164/re.2019.0320.
[21] B. K. P. Horn, Doubling the accuracy of indoor positioning: frequency diversity, Sensors
20 (2020) 1489. doi:10.3390/s20051489.
[22] B. K. P. Horn, Observation model for indoor positioning, Sensors 20 (2020). doi:10.3390/
s20144027.
[23] Y. Wang, C. Xiu, X. Zhang, D. Yang, WiFi indoor localization with CSI fingerprinting-based
random forest, Sensors 18 (2018). doi:10.3390/s18092869.
[24] F. DΓΌmbgen, C. Oeschger, M. KolundΕΎija, A. Scholefield, E. Girardin, J. Leuenberger, S. Ayer,
Multi-modal probabilistic indoor localization on a smartphone, in: 2019 International
Conference on Indoor Positioning and Indoor Navigation (IPIN), 2019, pp. 1β8. doi:10.
1109/IPIN.2019.8911765.