Content uploaded by Su Min Kim
Author content
All content in this area was uploaded by Su Min Kim on Feb 12, 2017
Content may be subject to copyright.
Baseline Performance of LTE Positioning in 3GPP
3D MIMO Indoor User Scenarios
Henrik Ryd´en, Sara Modarres Razavi, Fredrik Gunnarsson, Su Min Kim, Meng Wang,
Yufei Blankenship, Asbj ¨orn Gr¨ovlen, ˚
Ake Busin
Ericsson Research
Emails: (henrik.a.ryden, sara.modarres.razavi, fredrik.gunnarsson, su.min.kim, meng.a.wang
yufei.blankenship, asbjorn.grovlen, ake.busin)@ericsson.com
Abstract— Positioning in currently deployed LTE networks
is made via a combination of enhanced cell identity (E-CID),
observed time difference of arrival (OTDOA) and global navi-
gation satellite system (A-GNSS) information. Both simulations
and field trials indicate acceptable performance for outdoor
users. However, today the majority of the connections to cellular
networks are established from terminals located indoors. This
paper provides baseline positioning performance results based on
the 3GPP 3D MIMO deployment and propagation model that has
been adopted in the 3GPP Release (Rel.) 13 study item on indoor
positioning enhancements. Horizontal and vertical accuracies are
investigated for both outdoor-only and outdoor-indoor network
deployments.
I. INT RO DUCTI ON
Location-based services and emergency call positioning
drives the development of positioning in wireless networks.
Global navigation satellite system (GNSS)-enabled terminals
are capable of determining the position outdoors within some
meters, and a plethora of applications and services in termi-
nals take advantage of such accurate positioning. Positioning
support in 3GPP LTE (Third Generation Partnership Project
Long Term Evolution) was introduced in Rel. 9, and with some
improvements in subsequent releases, see [6]. This enables
operators to retrieve position information for location-based
services and to meet regulatory emergency call positioning
requirements in adequately deployed and synchronized net-
works. In currently deployed LTE networks, the User Equip-
ment (UE) position is determined based on a combination of
cell identity, OTDOA and A-GNSS information from the UE.
For further information about wireless network positioning in
general, and for LTE in particular, see [13], [15], [17].
The current combination of position information and report-
ing protocols give good accuracy for outdoor UEs in adequate
deployments [12], [17]. However, an increasing fraction of the
UEs use their cellular wireless network connection indoors.
Therefore, it is relevant to address indoor UE aspects of the
existing positioning support, and whether it is reasonable to
consider any enhancements. In particular, for indoor UEs the
horizontal position may not suffice, while also the vertical
position component is needed. In the United States, the Federal
Commission of Communication (FCC) have acknowledged the
challenge with indoor users, and proposed dedicated indoor
UE positioning requirements which also include a vertical
component. By 2020, the horizontal location of 70% of all
wireless 911 calls must be provided within 50 meters [1], [2].
Due to the increasing interest in positioning of indoor UEs,
3GPP has initiated a study item in Rel. 13 on indoor posi-
tioning enhancements [3]. The study will define 3D scenarios
with different combinations of deployed macro and small cells
(also referred to as micro/pico cells) to enable evaluations of
both the baseline performance of the current combination of
cell identity (CID) and OTDOA, as well as different potential
enhancements. The scenarios feature building models and 3D
propagation models to describe the challenging non-line-of-
sight (NLOS) and penetration loss effects.
The paper is organized as follows. Section II introduces
the baseline positioning support in LTE, such as A-GNSS,
CID, OTDOA and UTDOA in detail. Section III describes
considered algorithms for time delay estimation and position
estimation. Section IV describes some key simulation scenar-
ios agreed upon in the Rel. 13 study item. Section V presents
the considered performance evaluation framework of OTDOA
in this study. Section VI analysis the numerical results for each
scenario. Section VII derives the conclusions of the work.
II. PO SIT ION IN G IN LTE
The positioning architecture in the LTE operates via two
positioning protocols: LTE positioning protocol (LPP) and
LPP Annex (LPPa). LPP is used for communication between
the network node enhanced-serving mobile location center
(E-SMLC) and a UE, while LPPa is the communication
protocol between an enhanced node B (eNB) and the E-SMLC.
Different entities may initiate the positioning. For example, in
case of emergency calls, the positioning request is sent by the
mobility management entity (MME) to the E-SMLC, which
initiates suitable communication via LPP and/or LPPa. More
information on LTE positioning architecture and protocols can
be found in [6], [15]. The following four sets of methods and
their combination are already supported in LTE networks and
are briefly described, however the focus of this paper is on
OTDOA performance.
A. Assisted-Global Navigation Satellite System
The retrieved satellite signal measurements (e.g. from
Galileo, GPS, GLONASS, BeiDou) are used for these UE-
based and UE-assisted methods. The A-GNSS methods are ca-
pable of estimating the UE’s position in outdoor environments
with accuracy of few meters, however their performance fails
in environments where it is difficult to receive weak satellite
signals, such as indoor and dense urban environments. There-
fore these methods are not explored for indoor-positioning
study item of Rel. 13 and hence not in this paper.
B. Cell Identity
In this network-based method, the position of the UE is
associated with the serving cell identity (CID). Knowledge
about the location of the serving eNB is utilized in E-SMLC
to proximate the UE’s crude position. The accuracy of the
proximity position is directly dependent on the cell coverage
area. The support of this method has been mandatory since
Rel. 8. In case of using more data such as RF measurements
from multiple cells, timing advance and Angle of Arrival978-1-4799-9858-6/15/$31.00 c
2015 IEEE
(AoA) measurements, an enhanced CID (E-CID) method can
be supported.
C. Observed Time Difference of Arrival
The Observed Time Difference Of Arrival (OTDOA) is a
UE-assisted method, in which the UE measures the time of
arrival (TOA) of specific positioning reference signals (PRS)
from multiple eNBs, and computes the relative differences.
These received signal time difference (RSTD) are quantized
and reported via LPP to the E-SMLC together with an accu-
racy assessment.
Based on known positions of eNBs and their mutual time
synchronization, it is possible for the E-SMLC to estimate
the UE position from the RSTD and covariance reports using
multilateration. The accuracy depends on the radio conditions
of the received signals, number of received signals as well
as the deployment, which means that it will vary spatially as
discussed in [13], [17].
D. Uplink Time Difference of Arrival
UTDOA, which utilizes the uplink TOA, is an alternative
method to OTDOA standardized in Rel. 11. The measurements
are based on Sounding Reference Signals (SRS). UTDOA is
not yet available in practice and the analysis of this method
has not been considered in this paper.
III. OTDOA POSI TI ON EST IMATI ON
One key component in the OTDOA position estimation is
the PRS TOA estimation in the UE together with an accuracy
assessment of the TOA estimates. A second key component
is the positioning estimation in E-SMLC given RSTD reports
from the UE. This section describes in detail the considered
TOA and position estimation algorithms applied for OTDOA
implementation in this paper.
A. Time Delay Estimation
Here, we employ a threshold-based maximum-likelihood
(ML) TOA estimation, which detects the first tap delay by
choosing the earliest peak among multiple peaks larger than
a predetermined threshold value based on the correlation
between the received signal sequence and the transmitted
reference signal sequence.
In LTE, the QPSK-modulated PRS sequence is defined by
Zl,ns[m] = 1
√2(1 −2c[2m]) + j1
√2(1 −2c[2m+ 1]) ,(1)
where m= 0,1,...,2Nmax,DL
RB −1.Nmax,DL
RB is the max-
imum number of downlink resource blocks (RBs) allocated
for the PRS, nsis the slot number within a radioframe, lis
the orthogonal frequency division multiplex (OFDM) symbol
number within the slot, and c[·]denotes the pseudo-random
sequence generated by a length−31 Gold sequence [5].
According to the PRS mapping criterion in [5], the complex-
valued symbols are mapped to resource elements (REs). Let us
denote the mapped transmitted signal sequence in frequency
domain by Sl,ns[k]defined as
Sl,ns[k] = Zl,ns[m]if mis mapped to k,
0otherwise,(2)
where k= 0,...,Nfft −1and Nfft is the the size of
fast Fourier transform (FFT). An OFDM modulated reference
signal sequence in time domain after the inverse FFT (IFFT)
is given by
xl,ns[n] = rP
Nfft
Nfft−1
X
k=0
Sl,ns[k]ej2πkn
N,(3)
where Pis the eNB transmitted power, and n= 0,...,Nfft-1.
In an LTE system, consecutive PRS subframes (a.k.a. a
positioning occasion) are periodically transmitted downlink.
According to the standard [5], one positioning occasion may
contain up to six consecutive PRS subframes. The period of
one positioning occasion can be configured to every 160, 320,
640, and 1280 msec. The number of PRS occasions is a trade-
off between allocated resources for positioning on one hand
and cell hearability and positioning accuracy on the other.
In case each positioning occasion is configured with multiple
PRS subframes, the receiver can consider the received signal as
a concatenated long sequence that spans through all subframes
of the occasion.
Next, we consider the following tapped delay link channel
model to describe indoor multipath channels.
h(t) =
L−1
X
l=0
alδ(t−τl),(4)
where Lis the number of multipath taps, aldenotes the
amplitude of the l-th tap, τlindicates the time delay of the l-
th tap and δ(t)is the delta function, which is one when t= 0
and zero otherwise.
By going through the delay tapped channel, the received
signal sequence becomes
y[i] = h[i]∗x[i] + w[i],(5)
where the modulated OFDM symbols in (3), including the
cyclic prefix, creates the total transmitted signal sequence
x. The convolution operation is denoted by ∗and w[i]is
additive thermal noise at the receiver. To detect the first tap,
we compute the cross-correlation values between the received
and the reference sequence as follows:
R(τ) =
Nw−1
X
i=0
y[i]x∗[i−τ],(6)
where Nwis the search window for the positioning and
(·)∗denotes the complex conjugate. The cross correlation is
computed per subframe. In order to make use of multiple PRS
occasions, we combine the measurements accordingly
Rave(τ) = 1
|S|X
s∈S|Rs(τ)|,(7)
where Sis the set of cross correlation estimates per PRS
occasion. Finally, according to the ML criterion, the first tap
is estimated with the predetermined threshold value ζas
ˆτ= arg min
τRave(τ)
max{Rave}≥ζ.(8)
The threshold value has to be carefully chosen considering
the assumed multipath channels. In the simulation set-up of
this paper, we do not optimize the threshold value and use ζ=
0.5, which implies half of the strongest peak. Furthermore, we
add a constraint for the time estimate in which Rave(ˆτ)has
a negative derivative; this is to make sure that we have found
the top of the correlation peak.
The OTDOA requires the RSTD measurement of several
eNB at each UE. Let us denote the estimated TOA from the
i-th eNB to the UE by ˆτ(i). Then, the resulting RSTD between
eNB iand a reference eNB jis obtained by the difference
between the estimated TOAs times the quantization
ˆτ(i,j)=Qˆτ(i)−ˆτ(j),(9)
where the quantization resolution is 1Tsfor RSTDs within
±4096Ts, and 5Tsotherwise (Ts= 1/(15000 ×2048) is the
LTE basic time unit).
B. Positioning Estimation
Until now the OTDOA method for cellular networks applied
multilateration in two dimensions to produce horizontal accu-
racy, however we extend the formulation to the third dimension
in order to estimate the position in 3D space and hence adding
the vertical positioning estimation.
Denote the unknown three-dimensional (3D) UE’s position
by x= (x, y, z)T, and the set of eNBs in the network by
N={1, .., N }. The known location of each eNB i∈ N is
defined by li= (xi, yi, zi)T. The distance between the ith eNB
and the UE, ri, is computed as (10) based on the Euclidean
distance between the eNBs’ locations and the UE’s position.
ri=|x−li|=p(x−xi)2+ (y−yi)2+ (z−zi)2(10)
Based on OTDOA method, the RSTD measurements from
the UE to the network are used for the positioning computa-
tions, and the most Kpowerful eNBs out of Nare considered
for estimating the position of x.
Without loss of generality, we consider the reported RSTD
from (9) to be a relative distance report y(i,j)subject to an
additive error e(i,j). The error term thereby represents the TOA
measurement errors times the speed of light, as well as report
quantization and NLOS propagation effects.
y(i,j)=hOT DO A(x) + e(i,j)=ri−rj+e(i,j )(11)
To estimate the UE’s position, the challenge is to minimize
a given norm of the difference of actual measurements and the
measurement model. Here, we use Gauss-Newton algorithm.
The location of the serving eNB is set as the initial search
point. Each step is defined as in (12), while αkis the step
size. The correlation peak value (max{Rave }) for each link is
used to compute the Rmatrix. The method is similar to the
model in [16] where SNR values are used to construct the R
matrix elements. In general, high SNR should correspond to a
high correlation peak value. This is more explored in Section
V. The H(x) = ∇xhOT DO A(x)is explicitly driven for 3D
OTDOA in (13).
ˆ
xk=ˆ
xk−1+αk(HT(ˆ
xk−1)R−1H(ˆ
xk−1))−1
HT(ˆ
xk−1)R−1(y−hOT DOA (ˆ
xk−1)) (12)
H(x) =
x−x1
r1−x−x2
r2
y−y1
r1−y−y2
r2
z−z1
r1−z−z2
r2
x−x1
r1−x−x3
r3
y−y1
r1−y−y3
r3
z−z1
r1−z−z3
r3
.
.
..
.
..
.
.
x−x1
r1−x−xK
rK
y−y1
r1−y−yK
rK
z−z1
r1−z−zK
rK
(13)
TABLE I
SCE NAR I O PARAMETERS
Outdoor macro cell Outdoor small cell
System BW per carrier 10MHz 10MHz
Carrier frequency 2.0GHz 2.0GHz
Carrier Number 1 1
Total power per carrier 46dBm Outdoor: 30dBm
Antenna gain 17dBi 5dBi
Distance-dependent path Loss 3D-UMa[1] 3D-UMi[1]
Penetration for indoor UEs 20dB + 0.5din 20dB + 0.5din [2]
Shadowing/ Fast fading 3D-UMa[3] 3D-UMi[3]
Antenna pattern 3D [4] 2D Omni-directional
Antenna height 25m + α[5] 10m + β[5]
UE height 3(nfl −1) + 1.5m [6]
Antenna configuration 2Tx×2Rx
UE noise figure 9dB
UE speed 3 Km/h
[1] Referring to Table 7.2-1 in [10], [2] din: independent uniform random
value between [0,min(25,UE-to-cell distance)] for each link, [3] Refer-
ring to [9], [4] Referring to [9], [5] In our study: α∼uniform[-5,25], β∼
uniform[-5,10], [6] In our study: nfl ∈ {1,2,3,4,5,6,7,8}
IV. SIM ULATI ON STU DY
In this section, we thoroughly present the two key agreed
indoor positioning simulation scenarios in 3GPP Rel. 13. One
scenario assumes outdoor-only deployment of macro-only and
macro and small cells, while the other scenario combines the
deployment of outdoor macro cells with indoor small cells.
In addition, the study item also encompasses a scenario with
sparse random deployment of indoor nodes, which is not
addressed in this paper. The considered scenarios are expected
to have different performance in terms of hearability, quality
of RSTD measurements and possibility to apply 3D based
positioning. The 3D MIMO model, specified in Rel-12 [10],
which addresses 3D modeling has been considered as the
channel model for both scenarios. The full set of considered
scenario-parameters can be found in [22] and major ones for
the scenario set-ups in the current study for the outdoor-only
deployment are given in Table I.
A. Outdoor-only Deployment Scenario
In this scenario, an outdoor network deployment is consid-
ered for three different cases:
•Case 1: Macro + 10 small cells per cluster,
•Case 2: Macro + 4 small cells per cluster,
•Case 3: Macro-only.
The scenario consists of 7 macro sites each having 3 cells,
located in a hexagonal grid with inter site distance of 500 m. A
cluster is defined as a concentrated UE area, and the scenario
assumes one cluster per macro cell. In Case 1 and Case 2,
respectively, ten and four small cells are placed randomly
in each cluster. In this scenario, 2
3of the UEs are randomly
and uniformly dropped within the clusters and 1
3of the UEs
are randomly and uniformly dropped throughout the macro
geographical area. The UEs are also randomly and uniformly
dropped at eight floor levels where each floor height is equal
to 3 m. Fig. 1 illustrates the outdoor deployment scenario for
Case 2 in which the red dots are small cells and the blue dots
represent the UEs.
B. Outdoor-Indoor Deployment Scenario
While evaluating the OTDOA method for indoor UEs, one
valid assumption is to consider the presence of small cells
−800 −600 −400 −200 0 200 400 600 800
−600
−400
−200
0
200
400
600
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 19
20
21
Fig. 1. The outdoor deployment simulation scenario for Case 2
−800 −600 −400 −200 0 200 400 600 800
−600
−400
−200
0
200
400
600
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 19
20
21
Fig. 2. The indoor deployment simulation scenario
inside the indoor environments such as buildings. Hence, in
the second scenario, aside from the outdoor macro cells, there
are indoor small cells regularly deployed at each floor level of
each building. The UEs are randomly and uniformly dropped
at each floor of 4-floor height buildings. Fig. 2 illustrates
the indoor deployment scenario in which the green rectangles
represent the buildings which are randomly dropped in each
macro cell. All UEs are indoor in this scenario.
In this study, we use the radio distance wrapping technique
described in [19] in order to simulate a network in which all
sites are assumed to be surrounded by other sites. It means that,
there exists cells covering the UEs that seem out of coverage
in Fig. 1 and Fig. 2.
The configuration of the small cells at each floor level
is shown in Fig. 3. The reason for not deploying all the
small cells consequently in the hall ways, is to provide more
variation in their locations in terms of xy-plane. The detailed
scenario parameters for the indoor small cell deployment are
based on [20].
Fig. 3. Small cell locations in the buildings of the indoor deployment scenario
TABLE II
SCE NAR I O PARAMETERS
Parameter Value
Network synchronization Synchronized
Cyclic prefix Normal
Number of consecutive PRS subframes 1
Number of positioning occations 8
PRS periodicity 160 ms
PRS bandwidth 10 MHz
PRS muting on
Sampling frequency 30.72 MHz
αkin Gauss-Newton algorithm 0.05
V. PERF OR MAN CE EVAL UATI ON
In this section, the assumptions and evaluation framework
for OTDOA is described in detail. The assumptions for OT-
DOA and PRS characterization are summarized in Table II.
The receiver model described in Section III is evaluated by
simulating the TOA error translated into meters in the outdoor
deployment scenario Case 2. The threshold ζin (8) causes
a trade-off between robustness to noise and LOS detection.
Higher threshold can reduce the probability of a noise peak
detection, while decreasing the robustness in detecting weak
LOS paths. Fig. 4 shows the absolute TOA error for different
SNR values. Two thresholds are considered, ζ= 1 and ζ=
0.5, where ζ= 1 implies taking the strongest peak. The figure
shows that in case of ζ= 0.5for 90% of the measurements,
a link with SNR around -10 dB has TOA error less than 50
meters, while by changing ζ= 1, the TOA error is roughly
150 meters.
−30 −20 −10 0 10 20
0
100
200
300
400
500
SNR [dB]
TOA error [m]
50th − percentile using ζ =1
90th − percentile using ζ = 1
50th − percentile using ζ = 0.5
90th − percentile using ζ = 0.5
Fig. 4. Absolute TOA error in meters for different SNR in the outdoor
deployment scenario Case 2
This figure also illustrates the challenge in choosing a proper
threshold. One observation is that for low SNR region, by
taking the strongest peak, the TOA error is lower. Another
observation is that the TOA error statistics of links with
-10 dB SNR is very close to the ones with 10 dB SNR
and even higher. This shows that in high SNR region with
our considered receiver model the dominating error source is
mainly the NLOS condition. Note that based on this figure we
can conclude that choosing the receiver threshold based on the
SNR values can increase the accuracy of the TOA estimations.
We further study our chosen threshold (ζ= 0.5) by
examining the probability density function (PDF) of the TOA
error. Based on Fig. 4, we only consider links with SNR above
-10 dB in order to have TOA error less than 50 meters for 90%
of the measurements. The PDF of the TOA measurements of
the links with SNR above -10 dB is illustrated in Fig. 5.
−50 0 50 100 150 200
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
TOA error [m]
PDF
Receiver performance
Gaussian N(7,8.5)
Fig. 5. PDF for the TOA error in meters for links with SNR above -10 dB
using ζ= 0.5in the outdoor deployment scenario Case 2
Note that the error is not centered around zero, this can
be explained by that it is more probable to miss the LOS
peak than detect a peak before the LOS path, i.e. detecting
a noise peak. The objective with the fitted Gaussian in the
figure is to model the line of sight TOA error given the
considered receiver. Obviously, there are also contributions
from the NLOS propagation which also needs to be modeled
in order to get a complete error model, but it is also relevant
to single out the timing estimation error itself.
In LTE, an assisted data including the candidate cells for
measurements together with their PRS configuration of up to
24 neighbor cells is provided to the UE by the location server
[7]. Therefore, we consider up to K= 24 strongest neighbor
eNBs with SNR above -10 dB in the OTDOA method for each
UE.
It has been shown in [18] that applying muting offers a large
advantage in PRS hearability compared to no muting, therefore
we also expect that muting in time and frequency should be
applied in real deployments as much as possible. The PRS time
estimates for the simulation results in this paper are based
on muting interference scenario. This means that only the
desired PRS corrupted by additive thermal noise is considered.
The other PRS (i.e., those from other eNBs) are assumed to
be transmitted over orthogonal resources. Such interference-
free situation can be accomplished by PRS resource pattern
planning via physical-layer cell identity (PCID) planning,
and/or PRS subframe muting between different cell groups
[23].
The results are also based on a perfectly synchronized
network. In [24], we have shown that for indoor UEs, the
network synchronization error is a minor error component
compared to other factors, such as multipath and NLOS
propagation effects.
VI. NU MER ICAL RES ULTS
In this section the baseline performance of OTDOA and
CID are numerically evaluated for the presented scenarios. The
performance metrics are the cumulative distribution function
(CDF) of horizontal and vertical accuracy for indoor UEs.
0 20 40 60 80 100
0
0.2
0.4
0.6
0.8
1
Horizontal positioning error [m]
CDF
OTDOA − Case 1
OTDOA − Case 2
OTDOA − Case 3
Fig. 6. The OTDOA performance for horizontal position accuracy of outdoor-
only deployment scenario
0 5 10 15 20 25
0
0.2
0.4
0.6
0.8
1
Vertical positioning error [m]
CDF
OTDOA − Case 1
OTDOA − Case 2
OTDOA − Case 3
Fig. 7. The OTDOA performance for vertical position accuracy of outdoor-
only deployment scenario
Fig. 6 illustrates the CDF of the horizontal positioning accu-
racy of OTDOA with muting over all simulated indoor UEs for
the three outdoor deployment cases. The vertical dashed line
points out the 50 m horizontal positioning accuracy. Eight PRS
subframes are considered for RSTD estimation. The curves in
this figure show that first the position estimation of OTDOA
becomes more accurate with denser cell configuration, and
second the OTDOA with muting is able to give promising
positioning accuracy for indoor UEs even for Case 3 in which
we only have macro cells in the network.
Fig. 7 presents the same CDF curves for the vertical posi-
tioning accuracy of the three outdoor-only deployment cases.
The vertical dashed line points out to 3 m positioning error
which is equivalent to a floor height. The vertical accuracy
has almost a uniform distribution for all vertical errors. As
illustrated in Fig. 7, the vertical accuracy is not much better
than just guessing a floor level. We have limited our search
space to the buildings’ heights, therefore when increasing the
buildings’ heights, the vertical accuracy may degrade.
Table III gives numerical horizontal position estimation
accuracies for both OTDOA and CID. The values clearly
illustrate the OTDOA potential higher accuracy of the position
estimation in comparison to CID method.
Fig. 8 and Fig. 9 present the horizontal and vertical po-
sitioning estimation of CID and OTDOA with muting for
the outdoor-indoor deployment scenario, respectively. The
probability values of the horizontal positioning errors are also
given in Table III. The results show that both horizontal and
TABLE III
HOR IZ ON TAL A CC UR AC Y OF OTD OA AND C ID
Scenario Method 50%70%80%90%
error[m]
Outdoor-only, Case 1 OTDOA 13 19 25 36
Outdoor-only, Case 1 CID 42 86 122 184
Outdoor-only, Case 2 OTDOA 14 21 28 40
Outdoor-only, Case 2 CID 64 124 177 250
Outdoor-only, Case 3 OTDOA 17 27 36 60
Outdoor-only, Case 3 CID 212 268 308 404
Outdoor-indoor OTDOA 6 9 12 16
Outdoor-indoor CID 15 20 24 31
* The results are for 8-floor buildings.
0 10 20 30 40 50
0
0.2
0.4
0.6
0.8
1
Horizontal positioning error [m]
CDF
CID
OTDOA
Fig. 8. The OTDOA/CID performance for horizontal position accuracy of
the outdoor-indoor deployment scenario
vertical accuracy benchmarks are fulfilled with a great safety
margin by both methods. The reason to have the vertical
accuracy of 99% of the UEs with CID method within 1m
error, is that all of these UEs are served by a cell in the
same floor, and the difference between the antenna and the UE
heights is 1 meter. The CID can also successfully estimate the
position of each UE with a high accuracy due to dense small
cell deployment. Moreover, we should mention that OTDOA
accuracy is not significantly higher than simple CID for dense
indoor small cell scenario since the error components due
to multipath, RSTD measurement and quantization errors are
scalable to the cell size, while it also requires stronger post
processing.
VII. C ONCLUSION
The baseline performance of current LTE positioning meth-
ods for indoor environments have been investigated with re-
spect to both horizontal and vertical positioning accuracy. Two
3D MIMO scenarios from the Rel. 13 study item were consid-
ered for the simulations. Given an outdoor deployment with a
mix of macro and small cells, while considering muting, the
baseline performance of OTDOA in the considered scenario
with indoor users is good and meets the FCC requirements for
horizontal accuracy. In addition, with an indoor deployment
of small cells, positioning based on CID offers surprisingly
good performance in the considered scenario for horizontal
accuracy, while also providing a floor level vertical accuracy.
The results indicate that the increasingly wide deployment
of small cells indoor and outdoor make 3GPP positioning
technologies very effective in indoor UE positioning without
adding extra cost.
0 5 10 15
0
0.2
0.4
0.6
0.8
1
Vertical positioning error [m]
CDF
OTDOA
CID
Fig. 9. The OTDOA/CID performance for vertical position accuracy in the
outdoor-indoor deployment scenario
ACKNOWL EDG MEN T
The authors would like to thank Ali Zaidi and Jessica
¨
Ostergaard for their thorough review.
REF ERE NC ES
[1] FCC 14-13, Wireless E911 Location Accuracy Requirements, Third
Further Notice of Proposed Rulemaking, February 2014.
[2] FCC 15-9, Wireless E911 Location Accuracy Requirements, Fourth report
and order, January 2015.
[3] 3GPP RP-141002, Revised SID: Study on Indoor Positioning Enhance-
ments for UTRA and LTE, September 2014.
[4] 3GPP TS 36.133, Requirement for support of radio resource management,
Release 12.
[5] 3GPP TS 36.211, Physical channels and modulation, Release 12.
[6] 3GPP TR 36.305, Stage 2 functional specification of User Equipment
(UE) positioning in E-UTRAN.
[7] 3GPP TS 36.355, Evolved Universal Terrestrial Radio Access (E-UTRA);
LTE Positioning Protocol (LPP), Release 12.
[8] 3GPP TR 36.814, Further advancements for E-UTRA physical layer
aspects, Release 9.
[9] 3GPP TR 36.819, Coordinated multi-point operation for LTE physical
layer aspects, Release 11.
[10] 3GPP TR 36.873, Study on 3D channel model for LTE, Release 12.
[11] J.A. Del Peral-Rosado, J.A. Lopez-Salcedo, G. Seco-Granados,
F. Zanier, and M. Crisci, Achievable localization accuracy of the po-
sitioning reference signal of 3GPP LTE, in Proc. International Conf. on
Localization and GNSS (ICL-GNSS), June 2012.
[12] S. Fischer, Observed time difference of arrival (OTDOA) positioning in
3GPP LTE, Qualcomm Technologies, Inc., 2014.
[13] F. Gustafsson and F. Gunnarsson, Mobile positioning using wireless
networks, IEEE Signal Processing Magazine, July 2005.
[14] F. Gustafsson, Statistical Sensor Fusion, Studentlitteratur, 2012.
[15] A. Kangas, I. Siomina and T. Wigren, Positioning in LTE, in Handbook
of Position Location: Theory, Practice, and Advances, R. Zekavat, R. M.
Buehrer eds., Wiley 2011.
[16] R. Kaune, J. Hrst and W. Koch, Accuracy analysis for TDOA localization
in sensor networks, in Proc. International Conf. on Information Fusion,
2011.
[17] J. Medbo, I. Siomina, A. Kangas, J. Furuskog, Propagation channel
impact on LTE positioning accuracy: A study based on real measurements
of observed time difference of arrival, in Proc. PIMRC 2009.
[18] R. Srinivasan, G. A. Marshall, G. R. Opshaug, B. Ristic, K. A.
Burroughs, S. Fischer, S. Zhang, D. Henriksson A and F. Cardinal, PRS
Muting Pattern Assignment to Optimize RSTD Measurement Acquisition
for OTDOA Positioning in 3GPP LTE, International Technical Meeting,
January 2015.
[19] R1-135767, Initial calibration results for 3D channel model, 3GPP,
Ericsson, November 2013.
[20] R1-145396, WF on evaluation scenarios and assumptions for indoor
positioning study, 3GPP, November 2014.
[21] R1-150266, Preliminary simulation results for positioning of indoor
UEs, 3GPP, Ericsson, February 2015.
[22] R1-150856, Study on indoor positioning enhancements for UTRA and
LTE, 3GPP TSG RAN WG1, February 2015.
[23] R1-151942, Baseline evaluation results of OTDOA, 3GPP, Ericsson,
April 2015.
[24] R1-151943, Performance impact from imperfect network synchroniza-
tion, 3GPP, Ericsson, April 2015.
