Journal of Global Positioning Systems (2004)
Vol. 3, No. 1-2: 2-11
GNSS Indoor Location Technologies
Gérard Lachapelle
Position, Location And Navigation Group, Department of Geomatics Engineering, University of Calgary, Canada
Tel: [phone] E-mail: [e-mail]
Received: 15 November 2004 / Accepted: 3 February 2005
Abstract. This paper presents an overview of GNSS-
based indoor location technologies. Current and
emerging users and their potential requirements are
first discussed. Signal attenuation and multipath
caused under indoor environments are described. The
basic method to acquire and track attenuated signals,
namely longer integration of signal measurements, is
summarized. The need for assisted GPS is addressed.
Availability and accuracy performance currently
achievable under various conditions (wooden structure
building, single family residence, large sport facility)
are illustrated through selected test results. The
limitations of current technologies and potential
enhancements are discussed. These include
measurement noise, existing signal structure and future
enhancements, frequency and time errors, user motion,
sensor aiding such as ultra-tight integration, and
solution reliability and continuity. The paper
concludes with a discussion of receiver testing
standards. The possibility of using a GNSS hardware
simulator to create reproducible indoor environments
in order to overcome the controllability issue
encountered with real environments is analysed.
Key words: Indoor location, AGPS, HSGPS, aided
GPS, Indoor GPS
1 Introduction
The need for indoor location was initially spurred by
the mid-90s U.S. FCC decision to require mobile
phone service providers to locate emergency (E911)
callers with an accuracy requirement (October 1999
revision) of 100 m (67%) or 300 m (95%) for network-
based solutions and 50 m (67%) or 150 m (95%) for
handset-based solutions, the latter being the case with
the use of GPS. The most promising technologies to
achieve this on a continental level were cellular phone
network-based TDOA methods and GPS if the latter
could be made to operate under attenuated signal
conditions such as urban canyons, forested areas and
indoors. Attenuated signal environments are now often
simply labelled “indoor” environments for the sake of
simplicity. Early developments and testing of cellular
phone network-based TDOA methods resulted in
promising results with the use of GPS to precisely time
synchronize signal transmissions (e.g. Klukas et al
1997, 1998). However, the additional cell equipment
required proved to be a challenge. In addition, TDOA
being an hyperbolic location method (similar to Loran-
C for instance), observability was found to be low in a
urban environment given the rapidly changing
geometry of the available cells. Outside of urban areas,
the cell geometry was simply not present to meet
availability requirements. In parallel to the above
developments, experiments to increase the integration
time to potentially allow GPS signal reception under
attenuated signal environments yielded promising
results, especially with the use of assisted methods
(Petersen et al 1997, Moeglein & Krasner 1998, Garin
et al 1999). Given the existing space-based
infrastructure and coverage advantages of GPS that
makes possible an in-mobile phone solution, R&D
efforts on improving GPS performance intensified
rapidly.
It soon became obvious that continuous outdoor and
indoor location availability could be used in a large
number of applications such as personal digital
assistant location, asset tracking, vehicular navigation,
and no doubt many others to be discovered, in addition
to emergency services. These applications now form
part of a location-based services business that is
expected to grow from USD 0.5 B in 2003 to USD 28
B in 2008. One of the most important technical
questions arising is what are the performance levels
users want. Standard location and navigation
performance parameters are important, namely
availability, accuracy, reliability and integrity, and

Lachapelle: GNSS Indoor Location Technologies 3
continuity (outdoor-to-indoor and indoor). The
performance level for each of these parameters varies
and, except for the case of E911, has not been
rigorously defined yet, except as a wish list. A safe
assumption is the higher, the better! A wish often
expressed is “10 m in 10 s anywhere at any time”.
2 Signal Attenuation and Integration Issues
Signal attenuation occurs in any medium other than
free space. The atmosphere for instance attenuates the
GPS signals by about 1 dB under normal
circumstances. Signal attenuation caused by specific
materials can be determined experimentally under
controlled conditions (e.g. Klukas et al 2004). Gyprock
and plywood for instance attenuate GPS signals by
about 0.5 dB and 2.3 dB, respectively, while a cinder
block causes attenuation ten times higher, at about 23
dB. Attenuation due to the forestry canopy is highly
variable and depends on leaf size and thickness,
moisture content, forest density and trunk size. Steel,
reflective glass and other construction materials have
yet other signal attenuation properties. The problem in
characterizing signal attenuation and reflection indoors
is the presence and arrangement of diverse materials.
In situ measurements are preferable if a sufficiently
sensitive receiver is available in the first place. If large
attenuation and non-homogeneities occur, the signals
measured by the receivers might be echo-only signals
that may contain large errors, depending on indoor
geometry, as illustrated in Figure 1. Thus, indoor
location accuracy is not only a function of signal
attenuation but also of building geometry, even when
assuming fairly constant satellite geometry. This will
be further illustrated through examples later. In
addition to the level of attenuation and multipath, their
spatial and temporal variations is of interest and
importance as these affect position continuity.
Line-of-sight signals are already very week (around -
160 dBW or -130 dBm) due to the large free space
loss. This signal strength results in a SNR (signal-to-
noise ratio) of -18 dB when a 2-MHz front-end filter is
assumed. Signals are acquired and tracked by
correlation and integration, resulting in a processing
gain of 33 dB, if one assumes a pre-correlation
bandwidth of 2 MHz and an integration time of 2 ms.
This procedure increases the SNR to +15 dB which is
the approximate lower limit allowable for acquisition
(The lowest SNR required for tracking is
approximately at least 3 dB lower than the above, the
exact value depending upon the tracking loop
configuration). When the signal is attenuated,
integration can be done over a longer interval in order
to increase the processing gain. Coherent integration
can be done over intervals of up to 20 ms when bit
synchronization is achieved due to the limited length of
the GPS navigation message bits transmitted as part of
the signal. If integration over a longer period is
performed, a non-coherent technique that squares the
in-phase (I) and quadra-phase (Q) measurements has to
be used. Squaring also squares the noise, thus
diminishing the net gain of this technique (e.g.
Chansarkar & Garin 2000). The processing gain is
given by
Gtot = 10 log (Bpre x T) + 10 log (M) – Sqloss (1)
where Bpre is the pre-detection bandwidth, T is the
coherent integration time in ms, M is the number of
non-coherent accumulations, and SQloss is the
squaring loss due to non-coherent accumulation. If
coherent integration is extended over the full 20 ms,
the SNR gain, with respect to a 2-ms integration, is 10
dB. Coherent integration is often the only method used
in tracking mode. For signal acquisition, non-coherent
integration is often extended to a few to several
hundred ms to reach the minimum SNR value of 15
dB.
Transmitter
Reflected signal
ŅAttenuated
direct signalÓ
User Absorbed signal
Correct Range
Path of least resistance
Figure 1: Possible GPS signal propagation paths into a building
A basic assumption, when long integration is
considered, is that the Doppler error remains small or
constant during the integration period. However the
receiver frequency and time standard (FTS) drift and
unpredictable user antenna motion can cause
significant phase changes and can effectively limit the
integration time. The FTS error can be reduced by
using a higher grade unit. If only a better FTS is used
and the user antenna is in motion, separating antenna
motion and FTS drift will remain an issue. The use of
an inertial measurement unit (IMU) to measure antenna
motion will further improve performance, as will be
discussed later. The antenna gain pattern is also an
issue as the use of reflected signals will generally be
the norm indoor, in which case these signals will arrive
at the antenna from any direction. An antenna with a
high gain at low elevation is therefore preferable,
although this can result in a further loss (an ½
hemispherical antenna gain of 3 dB is usually assumed
for signal strength calculations). The problem will be

4 Journal of Global Positioning Systems
compounded if the antenna is subject to inversion as in
the case of a mobile phone.
2.1 Pseudorange accuracy versus signal attenuation
Under signal attenuation, the thermal noise
significantly affects pseudorange accuracy. This is
shown in Figure 2. The theoretical minimum
pseudorange standard deviation, calculated using the
Cramer-Rao lower bound for the BPSK C/A code is
shown for two pre-correlation bandwidths, namely 2
and 16 MHz. A DLL (Delay Lock Loop) loop
bandwidth of 0.3 Hz is assumed. Current HSGPS
receivers use a 2 MHz pre-correlation bandwidth.
C/No values of 24 and 8 dB-Hz correspond to
attenuation levels of about 20 and 36 dB, respectively.
The error growth as attenuation increases is indeed
significant, namely from 3 m to 115 m for the case of a
2-MHz pre-correlation bandwidth. The effect on
position accuracy would depend on the effect of
multipath that would likely dominate the error budget.
Figure 2 also shows the actual pseudorange standard
deviation as obtained using a SiRF receiver under a
signal attenuation range of 0 to 30 dB. The standard
deviation is actually better than that predicted by the
Cramer-Rao lower bound algorithm. This is likely due
to the use of a proprietary filter inside the receiver.
2.2 Signal acquisition and AGPS
Low pre-correlation SNR is one of the main issues for
acquisition. Indeed, acquisition requires a post-
correlation SNR value of around 14 dB, which is
higher than for tracking, for safe (low false alarm or
false detection) detection. Moreover, the acquisition
process has to deal with the fact that it has no a priori
knowledge of user motion, clock drift and bit
synchronization (limiting coherent integration) in a
standard standalone configuration. As a consequence,
more advanced techniques have to be used. The main
approaches to this problem are the use of large banks
of correlators (van Diggelen and Abraham 2001) and
assisted GPS (AGPS). The latter has become standard
for mobile phone location. A GPS reference station
provides various parameters to the user to reduce
TTFF, such as ephemeris, almanac, initial position,
reference time, Doppler information and DGPS
corrections (Syrjärinne 2001, Karunanayake et al 2004,
Weill et al 2004).
Figure 2: Pseudorange accuracy versus attenuation: Cramer-Rao
lower bound (2 MHz pre-correlation bandwidth, 0.3 Hz DLL
bandwidth), SiRF STARII measured accuracy
2.3 Using MEMS Inertial Measuring Units (IMUs)
An IMU consists of an assembly of gyros and
accelerometers deployed in a configuration suitable to
measure orientation change and acceleration in three
dimensions. The minimal configuration for a six
degree-of-freedom system consists of three units of
each type. A more complex configuration may have
redundant sensors on each axis (arrays of sensors) that
can potentially yield superior performance. They come
in a variety of performance classes, sizes, power
consumptions and costs. Any class can be used to aid
HSGPS. The better the class, the better the overall
performance of the combined system will be, although
no performance analysis of overall system
improvement versus IMU class and signal attenuation
is readily available. For low cost portable applications,
miniature MEMS IMUs are relevant. Miniature MEMS
gyros and accelerometers have currently dimensions of
no less than 10 x 10 x 5 mm and consume ≤ 10 mA. In
large quantities their prices are sufficiently low for
consideration. IMUs can be used for tracking loop
aiding in an ultra-tight integration mode, user motion
detection during signal integration and bridging user
positions during very short signal measurement gaps.
The bandwidth and noise reduction can result in a gain
of several dBs. They can at the same time be used in

Lachapelle: GNSS Indoor Location Technologies 5
parallel to increase position continuity and reliability as
discussed later. A higher position rate can be made
available if required.
There are however numerous challenges in effectively
integrating MEMS IMUs with HSGPS. They are much
affected by thermal effects, which is a major problem
for numerous applications. If temperature effects are
not compensated, a performance of the order of about 1
degree (or slightly better) per second is achievable. If
temperature effects are compensated using even a
simple linear model, a performance of better than 0.1
degree per second is expected. Another major issue
with using an IMU is the determination of the initial
attitude parameters. Also, the correlation of IMU
velocity errors with oscillator frequency drift has to be
dealt with. The algorithms and software will likely
have to be application specific in order to optimise
performance. The pay back is however potentially very
significant.
3 Indoor Positioning Performance Issues and
Examples
There are issues with all major positioning
performance parameters when using GPS indoor.
These can be summarized as follows:
Availability: The number of satellites available and
their geometry limit this performance parameter. Low
elevation satellites are not usually available indoor due
to excessive signal attenuation. Low availability
worsens geometry and reduces redundancy and
therefore the effectiveness of RAIM algorithms. The
use of miniature MEMS barometers to aid the height
component is very cost effective and improves not only
redundancy but also the horizontal dilution of precision
(HDOP). A miniature barometer, if operating in
differential mode, is sufficient to identify the relative
height of a user with an accuracy of about 2 m. Thus, a
change of floor can accurately be measured using only
one MEMS barometer. The use of a clock constraint,
if the latter is sufficiently accurate, also improves
redundancy. If both a clock constraint and barometry
are used, two satellites in a good geometry can deliver
a horizontal position solution. A third satellite will
enable a RAIM algorithm.
Accuracy: This performance parameter is affected by
high noise, echo-only or high multipath signals and
degraded geometry. This is why indoor GPS accuracy
is much lower than outdoor as will be seen in the
numerous examples described below. DGPS will help
more than in the outdoors because position accuracy is
total user range accuracy multiplied by the DOP. Since
the DOP is generally poor, removing the user range
error due to single point operation will help somewhat.
The use of the height and clock constraints described
above will also help. IMU aiding and filtering will
improve relative accuracy. Other aiding methods such
as pseudolites, UWB, cellular network TDOA
methods, will also help accuracy, in addition to the
other parameters.
Continuity: This is an especially serious problem
indoor due to the rapid temporal and spatial
decorrelation of multipath and the rapidly changing
satellite geometry as signals come in and out. Epoch-
by-epoch positions using an unconstrained least-
squares approach show large jumps. The use of a
Kalman filter with constraints adapted to the expected
user dynamics will go a long way in dealing with this
issue. The use of a self-contained low cost MEMS
IMU will further improve positioning continuity.
Reliability: The use of RAIM is effective insofar as
redundancy is available. Since redundancy is low
indoor, applying RAIM may decrease availability and
accuracy (Lachapelle et al 2004). Like in the case of
continuity, adding self-contained sensors, external
aiding and filtering with proper constraints will
improve the situation. The use of combined GPS and
Galileo will have a major impact on reliability
(Kuusniemi et al 2004a, b, Lachapelle et al 2004).
Indoor Location Examples
Several examples are described herein to illustrate
HSGPS performance under a variety of signal
attenuation conditions, from relatively low to high
signal attenuation building. Unless stated otherwise,
the position solutions were derived using an
unconstrained epoch-by-epoch least-squares algorithm
to better assess the true epoch-by-epoch effects of high
noise, multipath, satellite geometry and their temporal
variations.
Kinematic Positioning in a wooden building
The results of this test, conducted with SiRF receivers,
are shown in Figure 3. The building is a large barn
made of wood with a roof of asphalt shingles. The
attenuation was consistently below 10 dB. The data
was processed in differential mode to eliminate
atmospheric and orbital errors. A pedestrian walked
along the outer walls numerous times with the
equipment mounted in a backpack. A NovAtel
BlackDiamond™ GPS/INS system was used to provide
external reference positions (The pedestrian went
outside between every run to allow GPS to update the
tactical grade HG1700 INS unit. Each run was
completed in a few minutes. The epoch-by-epoch
least-squares horizontal positions, shown in Figure 3,

6 Journal of Global Positioning Systems
are accurate to about 5 m. An unconstrained least-
squares solution is used to allow a more realistic
performance analysis. In an operational environment, a
Kalman filter would yield smoother relative positions.
Availability with a HDOP of 4 or better is 95%. The
accuracy degrades significantly when the HDOP
exceeds 4. A similar test outdoor with similar HSGPS
equipment would yield a corresponding accuracy of
about 2 m. A standard receiver was used inside to
confirm that no signals could be measured. These
results therefore show the effectiveness of HSGPS in
this type of indoor environment.
Location in a N o rth American Residence
North American residences are typically constructed
with a wooden frame, gyprock inside and wood, stucco
or some other finishing outside. Depending on the
region, walls vary in thickness between 15 and 25 cm
and include insulating material. The roof material
varies. In the example reported herein, the roof consists
of concrete slates. A SiRF receiver was installed in the
garage located under the main living room. The
dimensions of the garage are approximately 8 x 12 x 3
m. The wooden garage door was kept closed during
the test. The results are shown in Figure 4. The receiver
was initialised outside and kept outside for about 30
minutes. Signal attenuation indoor varied between 10
and 30 dB, the latter number being at the tracking
capability limit of the receiver (MacGougan et al
2002). Despite this high level of attenuation, the
horizontal position accuracy was of the order of 10 m
(RMS), with a maximum error of 30 m. This case
illustrates well the expected location performance of an
emergency caller inside such a building or for that
matter of an asset tracking GPS device. The accuracy
is high enough to identify the specific residence where
the device is located.
Figure 3: Kinematic test results inside a wooden building
Figure 4: Static test results inside a North American residence
Positioning in the Calgary Olympic Oval
This test case is fully described by Dao et al (2004).
The Calgary Olympic Oval recreational facility, shown
in Figure 5a together with the two-antenna test
equipment, was built for the 1988 Olympic Winter
Games and includes a 400 m speed skating track
bordered by a running track used for testing. The Oval
is 25 m high at its centre, surrounded by concrete walls
and a row of windows near the roof at a height of
approximately 20 m. Surveyed points were established
along the running track to evaluate the accuracy of
HSGPS-derived positions. The test reported herein
was conducted in kinematic mode. The passing times
of the pre-surveyed reference points were tagged by the
pedestrian conducting the test. Again HSGPS SiRF
receivers were used. The C/No values for the satellites
available during the test are shown in Figure 5a. Given
that a normal C/No under LOS conditions is 44 dB-Hz,
one observes that the attenuation ranged between 10
and 27 dB, about the same as in the case of the
residence test case described earlier. The test was
conducted with two systems, with an inter-antenna
distance of 1 m to assess spatial multipath
decorrelation and the impact of antenna diversity on
position estimation under such conditions. The results
of one test run are shown in Figure 5b, together with its

Lachapelle: GNSS Indoor Location Technologies 7
position performance statistics. The 2 DRMS accuracy
with one antenna (black dots) is 45 m while the
corresponding accuracy using the combined
measurements (blue and green dots) from both
antennas is about 25 m. Two important aspects of
these results are that (1) antenna diversity can
effectively average out some multipath effects in such
an environment and (2) position accuracy and signal
attenuation are only partly correlated, the impact of the
indoor geometry and signal reflectivity also playing
important roles.
Pedestrian
with the
test-
package
Figure 5a: Calgary Olympic Oval Kinematic test and signal
attenuation
Figure 5b: Calgary Olympic Oval Kinematic test results
Location in a commercial building
This test was conducted inside the PLAN laboratory,
located in the Calgary Centre for Innovative
Technology (CCIT), as shown in Figure 6a. Signal
attenuation, shown in Figure 6b, ranges from 15 to 30
dB. The epoch-by-epoch position scatter, also shown
in Figure 6b, is relatively large and the 2D RMS
accuracy is 75 m. The use of RAIM to identify
unreliable solutions is effective but of limited help due
to low redundancy, the user generally has the choice
with a position with low or unavailable reliability
measure or no position at all. Thanks to a rapid
temporal decorrelation of multipath, the use of a batch
type solution over short intervals (10 – 20 s) improves
the accuracy to 58 m and, not shown here, the
reliability.
Test location
Figure 6a: Office building test set-up
Figure 6b: Office building test results
Vehicular navigation in urban canyons
This example, reported by Mezentsev et al (2002) and
illustrated in Figure 7, shows very well the advantages
and limitations of HSGPS. A six km loop in

8 Journal of Global Positioning Systems
downtown Calgary was driven numerous times. That
reported here is representative of the average
performance. A standard receiver was used in addition
to a SiRF receiver to show the performance
differences. The UL (upper left) graph shows the
trajectory in green and the fixes obtained with the
standard receiver in blue. Availability is very low but
accuracy very high. The UR graph shows the epoch-
by-epoch least-squares fixes obtained with the HSGPS
units. Availability is high but accuracy is low. No
RAIM algorithm was applied and errors of hundreds of
metres occur, due to echo-only signals reflected from
buildings and, possibly, cross-correlation effects. The
LL graph shows the HSGPS heavily filtered solution,
which is generally quite good but benefits from the
straight segments of the trajectory. The LR graph
shows a solution (green) based on the integration of the
HSGPS unit raw measurements with an automotive
grade Murata rate gyro. In the latter case, availability is
100% and the accuracy is excellent, apart from a bias
growing to 50 m in the LR section of the trajectory.
Further augmentation by other vehicle’s components
such as the ABS would likely further improve accuracy
and would certainly improve reliability, although the
latter was not systematically analyzed in this example.
Figure 7: Vehicular navigation in urban canyons
Pedestrian navigation in urban canyons
This case is much more difficult than the vehicular
case due to the more unpredictable nature of a
pedestrian’s trajectory. A totally unpredictable turning
radius might be a good way to describe this case! In the
example shown in Figure 8 and conducted by
Mezentsev et al (2004), a full six degrees of freedom
MEMS IMU attached to the user’s waist is integrated
with a HSGPS receiver in pedestrian dead reckoning
(PDR) mode. The HSGPS receiver is a SiRF X-trac
unit. The prototype test PDR unit consists of three
gyros and three accelerometers built in a classical
perpendicular triad scheme that represents a full six
degrees of freedom low-cost IMU, as shown in Figure
8. The gyros used are Analog Devices ADXRS150
±150°/s single chip rate gyros that are well suited for
pedestrian navigation due to their small size (7 × 7 × 3
mm) and low power consumption (< 50 mW). The
MEMS accelerometers used are VTI SCA 610 series.
The total volume including a power regulation circuit
is less than 100 cm3. In a commercial product, the
volume of such unit would likely be of the order of a
few cm3.

Lachapelle: GNSS Indoor Location Technologies 9
Since the IMU is attached to the body in a “semi-rigid”
manner, its measured dynamics do indeed represent
those of the user, which is not the case if the IMU was
mounted in a PDA or mobile phone. The gyro triad is
used to keep track of attitude. It is a complete tri-axis
quaternion based attitude solution, so user's heading,
pitch and roll are known. The accelerometers are used
to detect steps. Also, optionally, they can be used to
perform horizontal alignment at complete user stops.
The user step length is initially assumed to be constant,
say 70 cm. During use, it is calibrated when good GPS
position solutions are available. In PDR mechanization,
the position error is proportional to the distance
traveled and not time as is the case in a classical INS
mechanization. Given the position error growth
characteristics of both approaches, the former (PDR)
for a pedestrian significantly outperforms the latter,
even with tactical grade systems over long periods of
time (more than 1 minute of INS only navigation). The
GPS-IMU PDR integration steps used are shown in
Figure 8, together with the results of a 1.5 km loop in
downtown Calgary. A RAIM algorithm was
implemented to detect unreliable solutions. The
unaided least-squares GPS position fixes are shown in
green (reliable), red (Global test failure), and black (not
enough redundancy to judge). The integrated HSGPS-
IMU trajectory, obtained with a Kalman filter, is
shown in blue. Its availability is nearly 100% while its
maximum error reaches 50 m. An important question
is what the performance gain would be if an ultra-tight
integration was used.
SiRF HSGPS Unit
Low-Cost IMU-
belt mounted
Integration steps
Detect a step
Find heading
Estimate step length
PDR mechanization
LSQ + RAIM
HSGPS:
Pos/Vel correlation
INTEGRATION
Kalman Filter
Figure 8: Pedestrian navigation in urban canyons
4 Testing Procedures and Standards
The test results shown in the previous section illustrate
the difficulties associated with defining a sufficiently
wide range of “standard” environments to fully test a
commercial product purporting to meet pre-defined
minimum operational performance standards (MOPS).
Even if such environments were found, reproducibility
would be an issue due to the spatial decorrelation of
multipath, as was seen in the Calgary Olympic Oval for
instance. In addition such environments would hardly
be portable.
In order to establish common testing standards, the
question arises as to whether it is possible to construct
realistic environments on a simulator. If this were
possible it would resolve the above issues. An initial
attempt was made by Spirent with the development of
a new generation of simulators. Using its SimGEN
software, it is possible to simulate a wide range of
multipath scenarios, with different degrees of
obstructions up to echo-only signals, while varying
signal strength, as shown in Figure 9. The next step is
to “stochastically” reproduce a realistic environment.
This means simulating an environment with the same
stochastic signal fading, multipath and temporal
variation properties as those observed in an actual
environment. Early attempts have resulted in
encouraging results (Lachapelle et al 2003) but there is
still much work to accomplish to obtain satisfactory
results.
Simulated
Obstruction
Zone - Cat A
LOS-only
zone - Cat B
LOS+echoes
zone - Cat C
Echoes-
only zone -
Cat D
Spirent Simulator Capability
Figure 9: Indoor signal simulation scenario
5 Future Signals
GPS II and Galileo signals will have significant
advantages over the current GPS C/A code modulated
L1 signal. The availability of pilot channels will allow
the use of a pure PLL (Phase Lock Loop) and will
result in more robust carrier phase tracking, avoiding
the squaring loss currently present due to the need of a
Costas loop.
BOC (Binary Offset Carrier) modulation on the Galileo
E5 and L1 (and GPS M-code) will improve mitigation
of thermal noise, multipath and narrow band
interference. Figure 10 shows the Cramer-Rao lower
bounds for pseudorange accuracy for BOC(1,1)

10 Journal of Global Positioning Systems
modulation (that will be used for Galileo L1) versus
BPSK(1) (currently used on the GPS L1 signal). The
Cramer-Rao lower bound shown only deals with
thermal noise effects. The standard deviation
difference between a BPSK(1) modulation using a 2-
MHz pre-correlation bandwidth and a BOC (1,1)
modulation using a 16-MHz bandwidth is nearly one
order of magnitude. How this will translate in actual
position accuracy in the presence of large multipath
and echo-only signals remains to be determined. It
should also be noted that BOC signal tracking could
lead to biased pseudorange measurement due to its
multi-peak auto-correlation function. This might
particularly be a threat when low SNR are considered.
Finally the use of secondary codes (on GPS L5 and
Galileo L1, E5) will result in better narrow band
interference mitigation and improved bit
synchronization, although they might degrade the
acquisition MTTF (Hegarty et al 2003; Macabiau et al
2003).
Figure 10: Cramer-Rao lower bounds – BPSK(1) versus BOC(1,1)
signal modulation techniques
6 Conclusion s
The past 10 years have seen the birth, development and
deployment of the first generation of indoor GPS
technology. The current limitations of this technology
are severe, as compared to the level of performance
achievable to outdoor users. Yet, the technology is
largely responsible for the emerging location-based
services market, which has enormous potential for
growth and impact. This market is becoming
increasingly demanding in terms of performance.
Accuracy is highly addictive! The question now is
what will the technological improvements be during
the next 10 years. One can safely predict significant to
major improvements in the following areas: better
signal tracking, use of new GPS and Galileo signals,
and effective use of self-contained sensors and external
aiding. Will these improvements be sufficient to keep
up with users expectations?
Acknowledgements
The assistance of the following researchers in the
PLAN Group, Department of Geomatics Engineering,
the University of Calgary, are acknowledged: Drs. C.
Ma and M. Petovello, senior research associates, and
D. Dao, O. Julien, D. Karunanayake, H. Kuusniemi, O.
Mezentsev, and B. Zheng, MSc and PhD candidates.
References
Chansarkar, M. and L. J. Garin (2000): Acquisition of GPS
Signals at Very LowSignal to Noise Ratio. In
Proceedings of the Institute of Navigation ION National
Technical Meeting-2000 (January 26-28, 2000,
Anaheim, California), pages 731–737.
Dao, D., H. Kuusniemi, and G. Lachapelle (2004): HSGPS
Reliability Enhancements Using aTwin-Antenna
System. CD-ROM Proceedings of GNSS 2004
Conference, Rotterdam (16-19 May), 11 pages.
Garin, L. J., M. Chansarkar, S. Miocinovic, C. Norman, and
D. Hilgenberg (1999): Wireless Assisted GPS-SiRF
Architecture and Field Test Results. In Proceedings of
the Institute of Navigation ION GPS-99 (September 14-
17, 1999, Nashville, Tennessee), pages 489–497.
Hegarty, C., M. Tran, and A.J. Van Dierendonck (2003):
Acquisition Algorithms for GPS L5, CD-ROM
Proceedings of the European Navigation Conference
GNSS, (Graz, Austria, April 22-), 14 pages.
Karunanayake, D., M.E. Cannon, G. Lachapelle, G. Cox
(2004): Evaluation of AGPS in Weak Signal
Environments Using a Hardware Simulator.
Proceedings of GNSS 2004 (Session A6, Long Beach,
CA, 21-24 September), The Institute of Navigation,
Fairfax, VA
Klukas, R., G. Lachapelle, and M. Fattouche (1998): Cellular
Telephone Positioning Using GPS Time
Synchronization. GPS World Magazine, 9, 4, 49-54.
Klukas, R., G. Lachapelle, and M. Fattouche (1997): Field
Tests of a Cellular Telephone Positioning System.
Presented at IEEE VTC '97 - 47th Annual International
Vehicular Technology Conference, Phoenix, AZ, May
5-7.
Klukas, R., O. Julien, L. Dong, M.E. Cannon, and G.
Lachapelle (2004): Effects of Building Materials on
UHF Ranging Signals. GPS Solutions, 8, 1, 1-8.
Kuusniemi, H., G. Lachapelle, and J. Takala (2004a):
Reliability in Personal Positioning. CD-ROM
Proceedings of GNSS 2004 Conference, Rotterdam (16-
19 May), 14 pages.

Lachapelle: GNSS Indoor Location Technologies 11
Kuusniemi, H., and G. Lachapelle (2004b): GNSS Signal
Reliability Testing in Urban and Indoor
Environments. Proceedings of NTM 2004 (San Diego,
January 24-28), 210-224.
Lachapelle, G., H. Kuusniemi, D. Dao, G. MacGougan, and
M.E. Cannon (2004): HSGPS Signal Analysis and
Performance Under Various Indoor Conditions.
Navigation, U.S. Institute of Navigation, 51, 1, 29-43.
Lachapelle, G., M.E. Cannon, R. Klukas, S. Singh, R.
Watson, P. Boulton, A. Read and K. Jones (2003):
Hardware Simulator Models and Methodologies for
Controlled Performance Assessment of High
Sensitivity AGPS Receivers. CD-ROM Proceedings of
GNSS 2003, The European navigation Conference
(Graz, Austria, 22-25 April), Session A2 on Indoor
Navigation, 21 pages.
Macabiau, C., L. Ries, F. Bastide, and J.-L. Issler (2003):
GPS L5 Receiver Implementation Issues, Proceedings
of the US Institute of Navigation GPS/GNSS (Portland,
OR, USA, Sept. 9-12), pp. 153-164.
MacGougan, G., G. Lachapelle, R. Klukas, K. Siu, L. Garin,
J. Shewfelt, and G. Cox (2002) Performance Analysis
of A Stand-Alone High Sensitivity Receiver. GPS
Solutions, Springer Verlag, 6, 3, 179-195.
Mezentsev, O., J. Collin, H. Kuusniemi, and G. Lachapelle
(2004): Accuracy Assessment of a High Sensitivity
GPS Based Pedestrian Navigation System Aided by
Low-Cost Sensors. Proceedings of the 11th Saint
Petersburg International Conference on Integrated
Navigation Systems, 24-26 May, 156-164.
Mezentsev , O., Y. Lu, G. Lachapelle, and R. Klukas (2002):
Vehicular Navigation in Urban Canyons Using a High
Sensitivity Receiver Augmented with a Low Cost
Sensor. Proceedings of GPS2002 (Session E1, Portland,
OR, 24-27 September), The Institute of Navigation,
Alexandria, VA, 363-369.
Moeglein, M. and N. Krasner (1998): An Introduction to
SnapTrack Server-Aided GPS Technology.
Proceedings of the Institute of Navigation ION GPS-98
(September 15-18, 1998, Nashville, Tennessee), pages
333–342.
Peterson, B., D. Bruckner, and S. Heye (1997): Measuring
GPS Signals Indoors. Proceedings of the Institute of
Navigation ION GPS-97 (September 16-19, 1997,
Kansas City, Missouri), pages 615–624.
Syrjärinne, J. (2001): Studies of Modern Techniques for
Personal Positioning, Publication 319. Phd, Tampere
University of Technology.
van Diggelen, F. and C. Abraham (2001): Indoor GPS, The
No-Chip Challenge. GPS World, 12(9), pages 50–58.
Weill, L, N. Kishimoto, S. Hirata, and K. Chin (2004): The
next generation of a super sensitive GPS system.
Proceedings of GNSS 2004 (Session A5, Long Beach,
CA, 21-24 September), The Institute of Navigation,
Fairfax, VA.