You are on page 1of 5

Indoor Positioning Using Femtocells

Varun Khaitan, Peerapol Tinnakornsrisuphap and Mehmet Yavuz


Qualcomm Inc.
5775 Morehouse Dr., San Diego, CA 92121 USA
E-mails: {vkhaitan, peerapol, myavuz}@qualcomm.com

Abstract The ability to locate oneself indoors on a map and
navigate to desired areas has a wide range of applications. GPS
and macro cell tower based positioning are not very effective
indoors due to poor signal quality or limited accuracy in location
estimation. We propose to solve this problem by deploying an
indoor network of femtocells and determining the position of a
3G mobile using the downlink and uplink signals. The
measurement of signals from a group of femtocells at the mobile
and the measurement of the mobiles transmitted signal at group
of femtocells is used for triangulation of the mobiles position. It
is shown that a combination of these methods with a cleverly
designed transmission schedule involving time orthogonalization
of the femtocell signals can enable accurate indoor positioning.

Keywords: femtocell, interference, indoor, position, triangulation.
I. INTRODUCTION
Indoor positioning and navigation can open up a large
ecosystem of applications. To name a few, consumers/workers
will be able to locate themselves and friends in public places,
find the path to points of interest and receive information on
services in their immediate vicinity. Satellite-based positioning
systems perform poorly indoors as the signals from the
satellites are too weak to be decoded. Traditional terrestrial-
based positioning techniques used in macro cellular
environments also may not yield satisfactory accuracy required
for indoor applications.
Femtocells are low power base stations (BS) that extend the
range of and offload traffic off conventional wide area network
base stations or macrocells. They offer multiple benefits to
subscribers and operators. To subscribers, they promise
excellent user experience (better coverage and higher data
throughput) and access to specialized femtozone applications.
For operators, traffic will be offloaded from macro cellular
network thus offsetting the need to deploy additional cell
towers to sustain the increase in mobile data usage. Overview
of UMTS (Universal Mobile Telecommunications Systems)
femtocell (HNB) architecture and cdma2000 femtocell
architecture can be found in [1].
A network of femtocells deployed in building offers a
promising solution for indoor positioning as the coverage of
each femtocell is small and a finer resolution may be achieved
via positioning techniques such as signal strength triangulation.
Furthermore, the use of femtocells can facilitate the positioning
of existing 3G mobiles with no modifications nor requiring
support from any additional radio technology. This can be a
very important and distinguishing application for femtocells.
In this paper, we discuss key methods to identify the
position of a mobile in an indoor femtocell network. We
provide detailed implementation of the methods as well as
potential shortcomings and how they can be overcome. The
paper is organized as follows. Section II provides an overview
of different positioning techniques used in current cellular
networks. Section III presents the two broad classes of
techniques which can be used for positioning using femtocells.
Sections IV and V describe techniques based on downlink and
uplink signal strengths respectively. Section VI describes the
simulation model used to evaluate the performance of these
methods. Sections VII and VIII present the performance results
and techniques to enhance the performance further.
II. BACKGROUND
A. Overview of Positioning Techniques
Signal Strength Triangulation and Fingerprinting is a
method where the location of a mobile is estimated by
obtaining a set of signal strength measurements from a group
of transmitters and matching this set, known as fingerprint,
against a database of measurements from a grid of points in
the coverage area. The database is generated either by
physically making measurements at a large number of
locations or by using a computational network planning tool.
In [2], such a scheme has been evaluated experimentally for an
outdoor cellular network and the measurements include signal
strength as well as timing advance. Similar approaches are
evaluated in [3] and [4] for a commercial cellular system and
an indoor environment of wireless LANs.
Advanced Forward Link Trilateration (AFLT) [5] is a
positioning technology that relies on a time difference of
arrival of signals from multiple base stations at the mobile. To
determine location, the mobile reports the measurements of
signals from nearby synchronized base stations back to the
network, which are then used to triangulate an approximate
location of the mobile.
Highly Detectable Pilots (HDP) is a standardized pilot for
cdma2000 1xEV-DO (Evolution - Data Optimized) air-
interface that is transmitted in a dedicated slot with either
fixed or pseudo-random reuse pattern among neighboring cells
[6]. This time multiplexed transmission allows for improved
hearability of pilot signals from multiple cells.
Observed Time Difference Of Arrival (OTDOA) [7] is a
standardized positioning method for UMTS where the
observed time difference of pilots between a pair of base
station signals at the mobile is used to calculate an estimate of
the position (as a hyperboloid) and optionally, the velocity of
the mobile. The mobile's position is determined by the
intersection of the hyperbola for at least two pairs of base
stations. OTDOA can be enhanced by performing the
measurement of the non-serving cell during Idle Period
978-1-4244-8327-3/11/$26.00 2011 IEEE

Downlink (OTDOA-IPDL) of the serving cell, yielding similar
benefits to HDP.
Uplink Time Difference Of Arrival (UTDOA) [7] is also a
standardized positioning method for UMTS where the
observed time difference is calculated between the mobile and
a pair of Location Measurement Units (LMUs). The observed
time difference is calculated by maximizing the correlation of
time-shifted received signals at the LMUs.

B. Femtocell Overview
A femtocell may be deployed in the same frequency
channel with the macrocell (co-channel deployment) or in a
separate channel that is not in use by the macrocell (dedicated
channel deployment). When a mobile comes in close
proximity of a femtocell, it detects the femtocell pilot/beacon
and makes a handoff from the macrocell. Mobiles that are
operating on the same channel with the femtocell detect the
pilot through a neighbor list pilot search. For mobiles on the
macro-only channels, handoff is enabled through transmission
of pilot beacons [8][9]. Alternatively, the mobile may
autonomously perform inter-frequency scan due to weak
macrocell pilot or proximity to the femtocell [10]. Once on the
femtocell, the set of pilots searched is controlled by the
femtocell network.
Interference management is a critical aspect for successful
deployment of femtocells. In order to provide best
performance, the femtocells need to transmit their pilots and
beacons with enough power to ensure satisfactory coverage
while limiting the interference caused to mobiles on
macrocells. In this analysis, the transmit power of femtocells
on all channels is assumed to be calibrated for best
performance as discussed in [12].
III. FEMTOCELL BASED POSITIONING METHODS
To locate a mobile device in a network of femtocells,
whose locations are known, we need to determine its position
relative to at least three femtocells to achieve successful
triangulation. The distance between the mobile and a femtocell
is estimated by calculating the signal propagation loss
(pathloss) between them or the time taken by the signal to
propagate from one point to the other.
A. Signal Strength Triangulation based methods:
The presence of various obstructions in indoor
environment (furniture, walls, etc.) makes it impossible to
have an accurate mathematical model describing the relation
between distance and pathloss [11]. We implement a solution
where we generate a database of pathloss at all locations via
ray-tracing simulation of detailed building interiors using
WinProp software tool. The pathloss at the users location can
then be matched against the database to estimate the position.
B. Time based methods
The signal propagation delay between femtocells and mobile,
although useful in calculating distance, is ineffective as
1) The indoor radio propagation is location-dependent,
comprises of severe multipath and is unlikely to have line-
of-sight between the transmitter and receiver [11].
2) The resolution of the time estimate for 3G mobiles is
insufficient for indoor applications. The accuracy for
OTDOA measurements during IPDL gap is 0.5 chip [13]
(which is equivalent to 39 meters uncertainty).
Therefore, we focus on the pathloss based method to allow
effective indoor positioning of existing 3G mobiles with
femtocells.
IV. POSITIONING BASED ON DOWNLINK SIGNAL
STRENGTH
The position of a mobile can be estimated by measuring
the strength of the received downlink signals at the mobile
from a group of femtocells. When a position fix is needed,
either the client in the mobile triggers the measurement or the
serving femtocell requests the mobile to report signal strength
measurements from all the visible femtocells. If the client is in
the mobile, it is assumed that the mobile has some application
layer protocols to communicate the measurement and obtain
the map and position fix with a server.
To perform the measurement, the mobile is assumed to be
in an active call. If the user is not active, a dummy call can be
initiated. The specific procedures for reporting the pilots are
slightly different for cdma2000 1x and UMTS.
In cdma2000, the serving femtocell requests the mobile to
send a PSMM (Pilot Strength Measurement Message), either
once or periodically. As part of the PSMM report, the mobile
sends the Ecp/Io of all femtocells which are measurable and
the total Io, where Ecp is the received signal strength of the
serving femtocell pilot and Io is the total received energy on
the serving femtocell frequency (as measured by the mobile).
The pathloss to each visible femtocell can then be calculated
using the femtocell transmit powers. A similar procedure,
Candidate Frequency Search is used to report beacon signal
strengths measured on the macro frequencies.
In UMTS, the serving femtocell requests the mobile to
send a Measurement Report Message (MRM) which contains
Ecp/Io and Ecp information.
In either case, the fingerprint is matched against the
database, containing pathloss values from all points in the
networks coverage region to all femtocells. The point with the
maximum likelihood is reported as the predicted location. This
procedure is described in detail in Section VI.
The positioning accuracy depends on the number of
femtocells whose downlink strength can be measured. A
mobile can detect the pilot from a femtocell only if the signal
to interference-plus-noise ratio for the pilot (SINR or Ecp/Io)
is above a detection threshold (typically around -20 dB). The
downlink from other femtocells and macrocells (if the
femtocell operates on a channel that is shared or adjacent to a
macrocell channel) are interferers and affect the triangulation.
When a mobile is close to a femtocell, the interference
generated by the serving cell to the non-serving cells is high
and the triangulation set is degenerated to one. It is assumed
that the communication is on a licensed spectrum and there is
no spurious interference. If the triangulation set is one, the
location of the serving femtocell is chosen to be the predicted
location which leads to inaccuracies. For obtaining an

effective position fix in the close vicinity of a femtocell,
enhancements to increase the number of visible signal sources
are needed. The following methods can be used to create time
orthogonalization of signals to avoid persistent interferences:
(i) Inter-frequency Beacon Transmission and (ii) Coordinated
Silence
A. Inter-frequency Beacon Transmission
Each femtocell may transmit its beacon pilot on different
frequency channels in a time division multiplexed manner.
Therefore, as measurements are made by the mobiles on the
channels at multiple instances, the mobile will now be able to
detect signals from different femtocells as all other interferers
are removed. The pathloss to these femtocells can then be
used as a fingerprint to determine the location of the mobile.
When the beacons are transmitted on the macro channel,
there is additional interference due to the macrocell signals.
The number of visible beacons would be maximized if the
beacons can be transmitted on a clean channel (e.g., no
macrocell transmission) for positioning purposes although that
may not always be practical.
The beacons are assumed to be transmitted one at a time,
i.e., the beacon measurements are spaced in time. This
introduces additional positioning inaccuracies if the user is
moving as successive measurements correspond to different
locations. For best performance, non-interfering beacons
should be grouped to transmit together in order to minimize
the time required for triangulation. Note that only two beacon
pilots other than the serving femtocell are needed for
successful triangulation as the mobile already has
measurement of the serving femtocell pilot on its traffic
channel.
B. Co-ordinated Silence Techniques
Techniques such as HDP and OTDOA-IPDL also help
create time orthogonalization of signals to avoid the problem
of strong interference from the serving femtocell. As a result,
the number of visible femtocell signals measurable by the
mobiles will increase. Unfortunately, these features are
optional in standards and are not widely supported in mobiles.
Therefore, femtocells need to also support alternative solutions
for mobiles that are not equipped with these features.
V. POSITIONING BASED ON UPLINK SIGNAL
STRENGTH
As an alternative or complementary to the downlink
methods, the position of a mobile can also be estimated by
measuring the strength of the mobile uplink pilot as received
at a group of femtocells. Since the transmit power of the
mobile is unknown and dynamic, the pathloss cannot be
estimated from this measurement. However, the difference of
the measured strength at two femtocells is equal to the
pathloss difference from the mobiles location to these
femtocells. Thus the difference in the pathloss values can be
used as a fingerprint. However, this reduces the triangulation
set size by one. For example, if four femtocells are visible,
then the fingerprint has three values only.
The proposed method is as follows: when a mobile position
is needed, the positioning server instructs all femtocells
around the serving femtocell to make uplink measurements for
the pilot of the mobile. Those femtocells which can sense the
mobile send the measured Ecp/Nt (ratio of the mobiles pilot
strength to the sum of all other signals on the channel) and Nt
values to the positioning server. The server calculates the
pathloss difference to a number of pairs of femtocells and
matches this against the database to predict the mobiles
position. Again, the accuracy of this method will be limited by
the number of femtocells which can sense the mobile and that
can be calculated analytically as follows.
As the mobile in active call will be power controlled by the
serving femtocell, its uplink signal cannot be measured at far
located femtocells. It can be shown that, assuming perfect
power control, a non serving cell can detect the mobile only if
its pathloss to the mobile satisfies the relation
PI
nonseiving
- PI
seiving
< _
E
cp
N
t
]
th
- S
th


where [
L
cp
N
t

th
is the target mobile SNR and S
th
is the
minimum SNR below which a femtocell cannot sense the
mobile signal. If the [
L
cp
N
t

th
is set to -18 dB and S
th
is -30 dB
(both are typical values), only those femtocells whose pathloss
to the mobile is within 12dB of the pathloss to the serving
femtocell can detect the mobile. This results in a small
triangulation set and thus poor positioning accuracy. The
performance can be improved by increasing the SNR target for
the mobile from -18 dB to -10 dB and which increases the
pathloss margin from 12 dB to 20 dB. The increase in the
target Ecp/Nt has to be carefully evaluated as it also increases
interference to the uplink of macrocells in co-channel
deployments. However, since the pathloss from a mobile to
serving femtocell is generally very small compared to typical
pathloss to the macrocell, the increase may be acceptable in
typical deployments. Further, the target Ecp/Nt can be
increased selectively on the mobile that is being positioned.
VI. SIMULATION MODEL
We have developed a detailed pathloss model of an
enterprise building using the WinProp software tool. Using
field measurements we have been able to verify the accuracy
of the modeling software to be within 3-5 dB of the actual
values.
A. Building model
An enterprise building, Qualcomm Research Center, has been
modeled in a detailed manner using building floor plans which
can be directly imported to the software. Care has been taken
while choosing the construction material, both internally and
externally, in order to accurately model the pathloss from the
femtocells to different areas of the building. The modeling has
been done on a very fine grid of size 0.2 meters.
B. Simulation Assumptions
We assume a dedicated frequency deployment where the
femtocells are on a clean channel. There are multiple macro

channels, one being adjacent to the femtocell channel, and
beacons are transmitted on all of them. For our analysis, the
macrocells and femtocells are assumed to be 50% loaded, (ie)
their power is 50% of their allowed maximum. It is also
assumed that 20% of the full power is allocated to the pilot for
both macrocells and femtocells.
Channel fading is simulated and added to the path loss. A
Rician fading model with K-factor = 1.5 is used to model
fading on the femtocell downlink and K-factor = 5.0 is used to
model fading on the macrocell downlink. Doppler effect is
modeled based on the user speed and traffic.
VII. PERFORMANCE RESULTS
The performances of the methods discussed earlier are
presented in this section along with their shortcomings and
suggested improvements.
A. Femtocell Downlink Measurement Method
To simulate the performance of this method, we generate the
database of Ecp/Io strength seen at all points on the grid from
all the femtocells on the floor. Please note that six femtocells
are required for complete coverage area (42000 sq ft) as the
recommended coverage area per femtocell is 6000-8000 sq ft
for indoor office environments [12]. For each position of the
user on the grid, the estimated position is obtained using the
steps in Figure 1. In this process, visible femtocells refer to the
femtocells whose Ecp/Io is above the detection threshold
which is assumed to be -20 dB.
Our field tests show that, due to fading and multipath, the
pilot strength of a femtocell at a particular point can vary
approximately uniformly over time in a range of 6 dB. Thus in
Step 4 the likelihood function is based on a uniform
distribution on the interval [-3, 3]. For example, if the mobile
reported pathloss of -90 dB to a femtocell, the mobile would
be equally likely to be on any point in the database that has a
pathloss value in [-93, -87] dB to the femtocell. The predicted
point in Step 5 is chosen to be the mean value of all the
equally likely points so as to minimize the expected root-
mean-square (RMS) value of the error.


Figure 1 Calculation of position using downlink signal triangulation

The RMS error in estimated position at each point on the
floor calculated as an average over 30 faded instances is
shown in Figure 2.

Figure 2 RMS error of predicted position using downlink
triangulation

The results show that the method is very accurate in regions
of overlapping coverage between multiple femtocells as more
femtocells are visible. In the regions close to a femtocell, only
one femtocell is in the visible set as it creates high interference
to all other femtocells. As the predicted point is the location of
the femtocell itself, this also explains how the error is low at
the femtocell location and increases as the actual mobile
location moves outward.
The efficacy of this method is compromised by the
hearability problem. This is illustrated in Figure 3 which
shows the number of visible femtocells at each location for a
dedicated femtocell deployment, the enterprise being located
close to a macrocell.

Figure 3 The number of visible femtocells at each location

As discussed in Section IV, the hearability problem can be
alleviated via time orthogonalization techniques. Figure 4
shows the number of pilots visible at different points for a
macrocell edge deployment (enterprise is at the edge of the
macrocell coverage) after orthogonalization technique is
applied, assuming a minimum detection threshold of -20 dB
Ecp/Io. The corresponding result for macrocell site indicates
good visibility on the building side close to the macrocell and
lower numbers on the other side.


Figure 4 Number of pilots visible at macrocell edge after
orthogonalization
B. Femtocell Uplink Measurement Method
The performance of the uplink method is simulated using a
slightly modified procedure from the downlink. As stated
earlier, the pathloss from a femtocell varies uniformly in a 6
dB interval around the actual value in field measurement. Thus
the difference in pathloss values from two femtocells follows a
triangular distribution, i.e., the convolution of two uniform
distributions, between +6dB and -6dB.
Upon receiving the measurements from femtocells where
the user can be sensed, the set of pathloss differences is
computed. At each point in the database, the likelihood of it
being the mobiles current location is calculated and point
which maximizes the likelihood function is chosen as the
predicted position. The accuracy of the method depends on the
number of femtocells that can measure the mobile pilot at
different points. The results in Figure 5 are generated using
target Ecp/Nt setpoint of -10 dB as discussed in Section V.
The performance with a setpoint of -18 dB is quite
unsatisfactory.

Figure 5 Visible set size for uplink triangulation
Overall the performance of the uplink method is worse than
downlink as the triangulation set is always smaller by one.
From Figure 3 and Figure 5, we see that when the mobile is
very close to the serving femtocell, both the downlink and
uplink measurement methods fail to triangulate the position.
VIII. CONCLUSIONS
In this paper we presented methods using femtocells to
provide accurate indoor positioning of mobiles. It was shown
that downlink signal strength based methods work well when
the mobile is not close to a femtocell. When near a femtocell,
the performance can be improved by time orthogonalization
techniques. On the uplink, the triangulation performance can
be improved by selectively raising the target setpoint of the
mobile. The proposed methods are to be tested in a real world
system to understand the accuracy in presence of actual fading
and measurement errors.
IX. REFERENCES
[1] J. Chen, P. Rauber, D.Singh, C. Sundarraman, P.
Tinnakornsrisuphap and M. Yavuz, Femtocells
Architecture & Network Aspects, available at
http://www.qualcomm.com/research/femtocells
[2] H. Laitinen, J. Lahteenmaki and T. Nordstrom,
Database Correlation Method for GSM Location,
IEEE Conf. on Veh. Technol., Vol 4., pp 2504-2508,
Spring 2001
[3] M. Hellebrandt and R. Mathar, Location Tracking
of Mobiles in Cellular Radio Networks, IEEE Trans.
Veh. Technol., Vol 48. No 5., pp 1558-1562, Sept.
1999
[4] Y. Chen and H. Kobayashi, Signal Strength based
Indoor Geolocation, in IEEE Intl. Conf. on Comm.
pp 436-439, May 2002
[5] 3GPP2 C.S0022, Position Determination Service
Standard for Dual Mode Spread Spectrum Systems
[6] Q. Wu, W. Zhao, P. Black, Y. Tokgoz, R. Padovani,
cdma2000 Highly Detectable Pilot, IEEE ICC
Workshops 2008
[7] 3GPP TS 25.305, User Equipment (UE) positioning
in Universal Terrestrial Radio Access Network
(UTRAN)
[8] P. Humblet, B. Raghothaman, A. Srinivas, S.
Balasubramanian, C. Patel, and M. Yavuz, System
Design of cdma2000 Femtocells, IEEE
Communications Magazine, Sept. 2009.
[9] F. Meshkati, Y. Jiang, L. Grokop, S. Nagaraja, M.
Yavuz, and S. Nanda, Mobility and Femtocell
Discovery in 3G UMTS Networks, available at
http://www.qualcomm.com/research/femtocells
[10] 3GPP TS25.367, Mobility procedures for Home
Node B (HNB) Overall description3GPP TS
25.133, Requirements for support of radio resource
management (FDD)
[11] K. Pahlavan, X. Li, and J. Makela, Indoor
Geolocation Science and Technology, IEEE
Communications Magazine, Feb. 2002
[12] C.Patel, L. Grokop, V. Chande, V. Khaitan, M.
Yavuz and S. Nanda, Femtocell and Beacon
Transmit Power Self-Calibration available at
http://www.qualcomm.com/research/femtocells
[13] 3GPP TS 25.133, Requirements for support of radio
resource management (FDD).

You might also like