=Paper=
{{Paper
|id=Vol-2498/short45
|storemode=property
|title=Indoor location for emergency responders using LTE D2D communications waveform
|pdfUrl=https://ceur-ws.org/Vol-2498/short45.pdf
|volume=Vol-2498
|authors=Howard Fan,Anusha Kilari,Krishna Vemuri,Isaac Daffron,Julian Wang,Vijayavignesh Ramamurthy
|dblpUrl=https://dblp.org/rec/conf/ipin/FanKVDWR19
}}
==Indoor location for emergency responders using LTE D2D communications waveform==
https://ceur-ws.org/Vol-2498/short45.pdf
Indoor Location for Emergency Responders Using LTE
D2D Communications Waveform
H. Howard Fan1, Anusha Kilari1, Krishna Karthik Vemuri1, Isaac Daffron1,
Julian Wang2, and Vijayavignesh Ramamurthy1
1Dept. of EECS,University of Cincinnati, Cincinnati, OH 45221, USA
2Dept. of Architectural Engineering, Penn State University, State College, PA 16801, USA
fanh@ucmail.uc.edu
Abstract. It is well known that GPS signals are very weak so they cannot be
received indoors. Without GPS, indoor location is a difficult problem, especially
for emergency first responders since a pre-installed WiFi or beacon transmitter
infrastructure may no longer be available in an emergency situation. On the other
hand, emergency first responders must carry radios for critical communication
needs in emergency response missions. In this work, we use the LTE Proximity
Services (ProSe) mode for Device-to-Device (D2D) direct communications and
the system information resource blocks of the LTE sidelink communication sig-
nals to measure time of arrivals (TOAs) among a few such D2D communication
devices that form an ad hoc wireless network, thereby providing indoor location
service that uses no infrastructure, no additional hardware device, and uses very
little communications payload bandwidth. This concept is tested by simulation
and by implementation on a software defined radio (SDR) network which is
demonstrated in a real-world indoor setting.
Keywords:Indoor location, indoor navigation, GPS-denied, LTE ProSe, First-
Net, emergency first responder communication, SDR network.
This work is funded by NIST Public Safety Innovation Acceleration Program under
NIST award # 70NANB17H171.
1 Introduction
Indoor location/navigation is a difficult problem due to lack of GPS signals. Many pos-
sible ways have been envisioned and tried, ranging from image-based to magnetism
based location. The most successful indoor location technology nowadays is perhaps
WiFi or beacon based, through a ubiquitous smart phone app. Because of the intricate
and highly variable indoor floor plans and wall layouts, indoor navigation using WiFi
or beacon signals would require installations of many beacons at specific locations due
to potential lack of line-of-sight and the rich multipath propagation environments. For
the first responder application, however, potential factors such as fire and smoke may
reduce the effectiveness of image based methods, electric outage or building damage
2
may disable WiFi or beacons, so that many if not all of the above solutions may not be
applicable.
With the advent of Release 12 of the 3GPP specifications, LTE devices now have
the capability to support Device-to-Device (D2D) communications enabling direct
mode operations and Proximity Services (ProSe) that allows the devices to detect each
other and communicate directly with one another without the aid of a cellular network
infrastructure. In this paper, we present an innovative time of arrivals (TOA) based
method to provide indoor location services using the readily available LTE ProSe D2D
mode via the sidelink of the LTE emergency first responder communication signals,
i.e., we “piggy-back” our location services onto the readily available communications
signals. Therefore this approach requires no prior indoor WiFi or beacon infrastructure,
no additional location service hardware, with the wireless communication signals that
are guaranteed to be available. This method can be readily integrated into an LTE com-
munications platform in software, satisfying both the mission critical communication
and location based service requirements with one device. In addition, since we use the
system information blocks of the LTE sidelink communication signals to measure
TOAs, very little overhead in communications payload bandwidth is required.
In addition, we developed a BIM-IL (Building Information Modeling based Indoor
Location) portal which serves as the front end to disparate indoor location technologies
and an abundance of smart building data; this would serve as a “one-stop shop” for
public safety indoor location purposes. The public safety community can utilize this
portal, without a need of specific software installations and skills in BIM and data pro-
cess, to import BIM files from local files or Web Services and then overlay the indoor
location positioning data onto smart building data so first responders can have all
needed information in one display.
Researchers have explored using the LTE waveform in a cellular infrastructure for
location and navigation, see e.g. [1] and the references contained therein. In those works
signals from several base stations (called eNodeB) need to be present with the locations
of the base stations known. The receiver measures its pseudoranges to the base stations
by the downlink signals. In this present paper, however, we use the D2D mode or the
sidelink signals, and assume much fewer known user locations only for the purpose of
anchoring relative locations of all involved users. No eNodeB is assumed available. All
involved users are equal as user equipment (UE). We use Ettus universal software radio
peripherals (USRPs) as SDRs to implement the UEs.
2 The dTDOA Method and Location Computation
There are many ways in providing user locations based on RF signals without GPS [2].
Using round-trip time of arrival (TOA) or time difference of arrival (TDOA) to estimate
locations are common and reliable. However, round-trip TOA requires a radio to return
its received measurement signal with a prescribed delay from its reception and requires
a dedicated communications protocol among the radios. The TDOA measurements re-
quire the receiving radios’ clocks to be synchronized, which in turn requires significant
bandwidth resources. Unlike GPS, most radio transceivers are not accurately
3
synchronized, making the TDOA based methods inaccurate unless labor/resource in-
tensive synchronization of all participating nodes are performed.
We utilize the dTDOA method [3, 4] that does not require tight time synchronization
or round-trip TOA measurements. The central idea of the dTDOA is to construct dif-
ferential TDOA rather than TDOA. During the ranging process, each sensor in the net-
work takes turns to transmit a ranging signal, and all other sensors listen and measure
TOAs based on their own clocks. When all sensors have finished transmitting the rang-
ing signals, each sensor then takes turn to pass the TOAs of the received ranging signals
to a central processing unit. No tight time constraints need to be imposed on any of
these transmissions. Due to space limitation, the dTDOA method is not described here.
The interested reader can refer to [3, 4] for details.
As discussed in [3], for high SNR we obtain N(N-3)/2 independent dTDOA equa-
tions with N users forming an ad hoc network. Since all locations are relative, subject
to rotation, shifting and reflection,we need to know ground truths of some user location
parameters in order to anchor all other locations relative to, say, a building. For exam-
ple, we need to designate some user(s) outside a building having access to GPS, or
designate some user(s) at known locations relative to a building layout. It is easy to see
that in a 2D case where each user location is determined by two parameters xi and yi,
we need to know 3 parameters to anchor all other user locations. Therefore, in 2D we
need to have at least 6 users in an ad hoc network to obtain 9 independent dTDOA
equations to compute all user location parameters. In a 3D case, we need to know 5
parameters to anchor all other user locations, so that we will need to have at least 8
users in an ad hoc network to obtain 20 independent dTDOA equations to compute all
user location parameters.
Once all dTDOA measurements/estimates are collected, we obtain a set of non-linear
dTDOA equations in terms of user locations. An iterative method can be used to solve
for a maximum likelihood solution, see [3] for more details.
3 The LTE Sidelink Waveform and TOA Estimation
Initial acquisition of transmitter information is obtained by synchronization at the re-
ceiver. The details of different types of synchronization signals are described in [6].
Two types of synchronization signals are used in the present implementation, they are
1) Sidelink Synchronization Signals (SLSS) for synchronization in time and frequency,
and 2) Master Information Block SL (MIB-SL) for additional information. In SLSS
there are Primary Sidelink Synchronization Signals (PSS) and Secondary Sidelink Syn-
chronization Signals (SSS). Since there are multiple USRPs, we assign an SLSS ID to
each USRP to distinguish among all estimated TOAs of different USRPs. Conse-
quently, the receiver would know based on the detected SLSS ID which USRP it is
estimating TOA from. On the other hand, MIB-SL is transmitted over the PSBCH and
carries information regarding a few parameters such as bandwidth and frame number.
The PSBCH is transmitted in the same subframe as the SLSS, indicated in the following
resource block (RB) diagram of Figure 1.
4
Fig. 1.Synchronization Subframe resource block structure of LTE SL air interface, 5 MHz mode.
Vertical coordinate: time, horizontal coordinate: frequency.
For any available LTE bandwidth (BW) setting such as 3MHz, 5MHz, 10MHz and
20MHz, the occupied bandwidth in the resource grid by PSS/SSS signals is constant
regardless of the LTE signal BW, always 1.4 MHz with an effective bandwidth of 1.08
MHz. This synchronization signal BW, although adequate for communication purpose,
is too narrow for TOA and location purposes resulting in poor TOA measurement and
estimation. The TOA signal should be selected to encompass as large a BW as possible
while keeping the pseudorandom properties which will make an effective signal for
correlation and precise time estimation. Such a signal is readily available in the MIB
block of the LTE waveform, such as the PSBCH resource block. In order to maximally
use the available BW for better TOA and location accuracy, we use the PSBCH signal
other than the PSS/SSS signals for TOA estimation, see Figure 1. The PSBCH signal
in the OFDM “frequency domain” (to be transmitted) turns out to be pseudorandom in
nature, therefore suitable for TOA timing estimation. In addition, since the MIB is a
control signal, using it for TOA estimation does not incur any additional BW overhead.
Table 1 lists the available LTE sidelink signal BWs and their corresponding
PSS/SSS and PSBCH resource block signal BWs, respectively. It is seen that the
PSBCH signal uses the available LTE sidelink waveform BW much more fully than
the PSS/SSS signals. The asynchronous TOAs are estimated by calculating the relative
time delay between two USRPs. Coarse time delays are estimated by calculating the
Sync point in a subframe obtained from correlation of synchronization (PSS/SSS) sig-
nals that lock at a time sample. Fine time delay estimation uses PSBCH signals for its
larger BW, and is performed by early-prompt-late correlators that achieve better time
resolution than the sampling interval.
While the TOAs are measured using the PSBCH signal which is part of the LTE
communication signal overhead to be transmitted with or without location service, the
estimated TOAs are transmitted to a processing center (one of the UEs) in the form of
payload data along with their corresponding IDs to distinguish between different TOAs.
5
Physical Sidelink Discovery Channel (PSDCH) in a payload subframe carries the user
data. Hence, we use PSDCH to transmit the needed TOA values, whose decoding at the
receiver is done normally as in [5]. OFDM demodulation and turbo decoding are per-
formed on PSDCH resource blocks in order to acquire the TOA values.
Table 1. LTE sidelink signal BW modes and signal BWs
LTE BW Effective Signal BW (MHz)
Mode (MHz) PSS/SSS PSBCH
1.4 1.08 1.08
3 1.08 2.7
5 1.08 4.5
10 1.08 9
15 1.08 13.5
20 1.08 18
4 A Multiuser Communications Protocol
There is no eNodeB in D2D communication, all devices are considered as UEs. Also,
there is only one carrier frequency in sidelink. Since multiple users need to be involved
in using the dTDOA method to compute user locations, a multiuser communications
protocol must be developed to ensure that only one UE transmits at any given time so
no over-the-air (OTA) collision will occur. A cooperative communications protocol
was proposed to avoid collision and achieve efficient communication among multiple
devices in LTE cellular mode of operation [7]. To our knowledge no multiuser cooper-
ative communications protocol exists today for ProSe D2D operation. In this section
we describe such a simple protocol that we developed for our purpose.
Such cooperative multiuser communication is achieved by giving unique IDs to all
UEs and providing different time slots for them to transmit and receive, see Figure 2.
User ID is attached to the Sidelink waveform by inserting the user ID into the payload
of transmitting waveform which is collected and decoded at the receiver at the time of
TOA measurement.
In this experiment six UEs are involved to find their 2D locations. Since all six user
IDs are assigned beforehand, a pre-assigned user Tx and Rx timing diagram is followed
as shown in Figure 2. For simplicity, all UEs start manually at approximately the same
time, then the timing of Figure 2 is followed by all UEs. A more sophisticated protocol
should take into consideration of UE discovery and participation into such a multiuser
network at random times. However, that is beyond the scope and budget of this project.
In real time implementation, manually starting at the “same time” does not result in
synchronization.It has been concluded from our USRP tests that, to ensure successful
reception, the transmitted signal (40ms of LTE SL subframes for all needed infor-
mation) should be repeated for a certain time period which is the time for each stage in
Figure 2. During this time period, the receiver UEs keep receiving the transmitted
waveform and 1) perform correlation to find TOAs and 2) decode the received
6
waveform to find the ID of the transmitting device and other TOAs in other stages. In
this experiment, six stages of transmission and reception are required to obtain all the
measured TOAs at the processing center. Details are omitted due to space limitation.
Stage 1 Stage 2 Stage 3 Stage 4
UE1 Transmit Discovery Waveform (TDW) Idle State Idle State Idle State
UE2 Receive and Measure TOA with UE1 TDW with TOA UE1-UE2 Receive and Measure TOA with UE3 Idle State
UE3 Receive and Measure TOA with UE1 Receive and Measure TOA with UE2 TDW with TOA UE1-UE3, UE2-UE3 Idle State
UE4 Receive and Measure TOA with UE1 Receive and Measure TOA with UE2 Receive and Measure TOA with UE3 TDW with TOA at UE4- UE1, UE2, UE3
UE5 Receive and Measure TOA with UE1 Receive and Measure TOA with UE2 Receive and Measure TOA with UE3 Receive and Measure TOA with UE4
UE6 Receive and Measure TOA with UE1 Receive and Measure TOA with UE2 Receive and Measure TOA with UE3 Receive and Measure TOA with UE4
Fig. 2.Multiuser D2D Protocol Timing Diagram (Only 4 stages are shown)
UE6 will be the processing center where all TOAs are collected and will be used to
calculate the locations by using the dTDOA algorithm. Once the locations are found, if
desired UE6 can send out the locations to all other UEs through payload, and a com-
munication cycle ends after this stage. If the TOAs are not available or not received for
UE6 to calculate locations, then this final stage transmission will not happen since lo-
cations cannot be calculated at the processing center. Since there are 6 UEs, there are
15 different combinations of TOAs by the end of one communication cycle. These 15
TOAs form 9 independent dTDOA equations.
5 Multipath Mitigation
Multipath propagation is a significant problem in indoor environments. Communica-
tion signals are easily reflected and attenuated by interior walls and equipment. This
causes signal distortion and erroneous TOA estimates at the receiver. While the LTE
standard includes OFDM techniques to tackle some issues of multipath, such as the
introduction of cyclic prefix, it was developed for efficient bandwidth use for commu-
nication, but not as a location-finding waveform for precise timing. In particular, the
basic pulse of the LTE waveform is “shaped”, unlike the GPS signals having unshaped
pulses. These shaped pulses make the correlation function look like sinc rather than
triangular, so multipath mitigation is especially challenging since the conventional cor-
relation-based methodswill have poor resolution with closely spaced paths.
Several line-of-sight (LOS) TOA estimation and multipath mitigation techniques
have been investigated, including some well-known methods in or outside the GPS lit-
erature. Due to the reasons mentioned in the previous paragraph, many of those meth-
ods do not work well for the LTE waveform, such as the slope differentiation method
and the MUSIC method. The projection onto convex sets (POCS) method works well
but is computationally intensive. Instead, we proposed a simple modified correlation
method (MCC) and tested it in both simulation and actual OTA field tests.
The MCC method sets a dynamic threshold which is empirically set at the average
plus three times the standard deviation of the received power of each frame. It then
7
selects the earliest peak above this threshold as the LOS TOA peak. Fine time delay
estimation is performed by early-prompt-late (EPL) correlators with a quadratic curve
fit that achieves better time resolution than the sampling interval. As discussed before,
we use PSBCH signal for better TOA estimation accuracy. A larger signal BW not only
results in better correlation resolution or tighter correlation peaks, it also allows multi-
path signals that are closely spaced in time to be more easily resolved.
6 Building Information Modeling – Indoor Location
Successful indoor location-based service (LBS) applications rely not only on indoor
maps, spatial data, and additional semantic building information, but also additional
dimensions regarding the operational performance of indoor devices, objects, and phys-
ical environments. The Building Information Modeling (BIM) software platform is now
widely used throughout the life cycle of a building from design and construction to
operation. These ready-to-use models are gradually replacing paper-based drawings
and pure geometry-based electronic models [8]. BIM has the potential to provide valu-
able data (e.g., real-time occupancy numbers, physical environment information, etc.)
to support first responders’ decision-making.
We developed a BIM-IL (Building Information Modeling – Indoor Localization)
portal which serves as the front end to indoor LBS and an abundance of smart building
data as a “one-stop shop” for public safety indoor location purposes. The public safety
community can utilize this portal, without a need of specific software installations and
skills in BIM and data process, to import BIM files from local files or web services and
then overlay the indoor location positioning data and smart building data, so first re-
sponders can have all needed information in one display. An existing BIM file can be
obtained from ownersor construction teams. The portal then automates the generation
of a BIM-based userinterface, including three main modules: simplified geometric data,
emergency-related semantic information, and a spatial data network for navigation.
Each module could be used separately or collectively for various public safety-based
indoor location purposes in terms of indoor positioning data integration, navigation and
mapping, and emergency-related semantics display. If BIM is not available, e.g., an old
building, then any available building layout or any additional information can still be
imported into the portal for integration with location data.
The methodology used to generate this interface is through the Unity software plat-
form which offers various visualization methods and potential immersive user experi-
ences via virtual reality. The BIM file is automatically converted to a .FBX file (con-
taining simplified geometry data) and a .CSV file (containing emergency-related se-
mantic data). The simplified geometric data include the main structures – walls, roofs,
ceilings, floors, windows, and stairs; and the emergency-related semantic data are re-
lated to fire rating hours, fire cabinet, room numbers, exit locations, hazard material
locations and types, etc. For example, semantic data visualization can use pseudo color
– walls with 3-hr fire rating are colored pink, and 2-hr fire rating red.
8
7 Real-Time OTA Test & Integration with BIM-IL
We performed field tests of the proposed system in an indoor setting and integrated
with BIM-IL portal. Six UEs are deployed using six USRPs, among them two are an-
chored, the other four UE locations are computed using the proposed system and meth-
ods. Shown below in Fig. 3 are a screenshot of BIM-IL portal with all six UE locations
indicated by colored icons (four are computed). The test room size is approximately
15m x 30m. The LTE signal BW is 5 MHz. Visual inspection against actual locations
indicate that 2D location errors are about a few meters.
Fig. 3. Left: Each USRP set (one UE) includes a laptop and antennas for additional gain; Right:
Screenshot of OTA indoor field test result integrated with BIM-IL portal, the white
colored dividing walls are absent in the test, the dark blue and purple icons are anchor
UEs, the light blue UE is purposefully set in an adjacent room separated by a wall so
line-of-sight path is completely blocked from all other UEs.
References
1. Shamaei, K. and Kassas, Z.M.: LTE receiver design and multipath analysis for navigation
in urban environments. NAVIGATION, Journal of The Institute of Navigation, 65(4), 655-
675 (2018).
2. Patwari, N., Ash, J.N., Kyperountas, S., Hero III, A.O., Moses, R.L. and Correal, N.S.: Lo-
cating the nodes “cooperative localization in wireless sensor networks”. IEEE Signal Pro-
cessing Magazine, 22(4), 54–69 (2005).
3. Fan, H. and Yan, C.: Asynchronous differential TDOA for sensor self-localization. Proc.
Intl. Conf. on Acoust., Speech, Signal Processing, pp. II–1109–II–1112 (2007).
4. Yan, C. and Fan, H: Asynchronous differential TDOA for non-GPS navigation using signals
of opportunity. Proc. Intl. Conf. on Acoust., Speech, Signal Processing (2008).
5. Gotsis, A., et. al.: Experimenting with Flexible D2D Communications in Current and Future
LTE networks: A D2D radio technology primer & software modem implementation. WInn-
Europe (2017).
6. Schlienz, J. and Roessler, A.: Device to Device Communication in LTE [White paper]:
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/application_notes/
1ma264/1MA264_0e_D2DComm.pdf, last accessed 2015/09/29.
7. Li, Y., Zhu, S. and Guo, X.: System Design for Multiple Users Cooperative Communication
in LTE. IEEE 78th Vehicular Technology Conference (VTC Fall), 1-5 (2013).
8. Pezeshki, Z. and Ivari, S.A.S.: Applications of BIM: a brief review and future outline. Ar-
chives of Computational Methods in Engineering, 2(25), 273-312 (2018).