INV ITEDP A P E R
Satellite Radiolocalization FromGPS to GNSS and Beyond: NovelTechnologies and Applicationsfor Civil Mass MarketThe current and forthcoming GNSS and associated technologies are discussed from
a mass-market perspective; hints are given about the future role of digital signal
processing and the software-defined ratio.
By Carles Fernandez-Prades, Member IEEE, Letizia Lo Presti, Member IEEE, and
Emanuela Falletti
ABSTRACT | It is known that satellite radiolocalization was
born in the military environment and was originally conceived
for defense purposes. Nevertheless, the commercial explosion
(dated to 20 years ago) of global positioning system (GPS) in the
civil market (automotive, tourism, etc.) significantly changed
the original perspectives of this technology. Another big
change is expected when other global navigation satellite
systems (GNSSs) such as the European Galileo or the Chinese
COMPASS become operational and commercial. In fact, modern
GNSSs are conceived principally for the civil market (at the
opposite of GPS, whose civil employment is given as a sort of
Bkind gift,[ with lower performance than that one granted to
military users). The scope of this paper is to provide readers
with a clear focus about the potentialities of current and forth-
coming GNSSs and associated technologies in a renewed mass-
market perspective. The paper also opens a window to the
future of radiolocalization technology beyond GPS and GNSS,
dealing with the role of digital signal processing and software-
defined radio (SDR) in next-generation navigation systems and
with the seamless integration of satellite-based navigation with
other technologies in order to provide reliable position infor-
mation also in hostile environments.
KEYWORDS | COMPASS; European Geostationary Navigation
Overlay Service (EGNOS); Galileo; global navigation satellite
systems (GNSSs); global positioning system (GPS); GLONASS;
Indian Regional Navigation Satellite System (IRNSS); inertial
navigation; Quasi-Zenith Satellite System (QZSS); radio naviga-
tion; Satellite-Based Augmentation System (SBAS); software
receivers; Wide-Area Augmentation System (WAAS)
I . INTRODUCTION
The art of finding the way from one place to another is
called navigation. Until the 20th century, the term re-
ferred mainly to guiding ships across the seas. Indeed, the
word Bnavigate[ comes from the Latin navis (meaning
Bship[) and agere (meaning Bto move or direct[). Today,
the word also encompasses the guidance of travel on land,
in the air, and in inner and outer space.
This encyclopedic definition talks about art, but naviga-tion has been (and today still is) a major scientific and
technological challenge. Navigation seems to start very
early in time: according to Chinese storytelling, the
Manuscript received October 13, 2010; revised March 14, 2011; accepted May 14, 2011.
Date of publication July 7, 2011; date of current version October 19, 2011. This work was
supported in part by the European Commission in the framework of the FP7 Network
of Excellence in Wireless COMmunications NEWCOM++ (Contract 216715) and COST
Action IC0803, and by the Spanish Science and Technology Commission under Project
NARRA (TEC2008-02685/TEC).
C. Fernandez-Prades is with the Communications Subsystems Area, Centre
Tecnologic de Telecomunicacions de Catalunya (CTTC), 08860 Barcelona, Spain
(e-mail: [emailprotected]).
L. L. Presti is with the Dipartimento di Elettronica, Politecnico di Torino, 10129 Torino,
Italy (e-mail: [emailprotected]).
E. Falletti is with the Navigation Signal Analysis and Simulation Group, Istituto
Superiore Mario Boella (ISMB), 10138 Torino, Italy (e-mail: [emailprotected]).
Digital Object Identifier: 10.1109/JPROC.2011.2158032
1882 Proceedings of the IEEE | Vol. 99, No. 11, November 2011 0018-9219/$26.00 �2011 IEEE
compass was discovered and used in wars during foggyweather before recorded history. Early mariners followed
landmarks visible on shore, until other technological
breakthroughs came into play: the magnetic compass
(c. 13th century), the astrolabe (c. 1484), the sextant
(1757), the use of lighthouses and buoys, or the seagoing
chronometer invented by Harrison in 1764 [1]. These
inventions fueled disciplines such as surveying and
geodesy, but also had a dramatic impact on transportation,and thus on economics. Today, the virtuous circle of
science, technology, and business around navigation is still
spinning and well alive, as will be described along the
following sections of this paper.
The application of radio waves to determine a position
starts in the 20th century with the patent of the first
direction finding system in 1902 [2]. World War II spurred
hyperbolic, ground-based low-frequency radio navigationsystems such as Decca and LORAN. Some time after,
technology allowed us to put radiophares in the sky.
Satellite-based navigation, which now is usually referred to
with the general framework of global navigation satellite
systems (GNSSs), started in the early 1960s with the
TRANSIT system, based on the fact that if the satellite’s
position were known and predictable, the Doppler shift
could be used to locate a receiver on Earth. The Cold Wararms race and associated military needs, mostly related to
ballistic missiles guidance and the nuclear deterrence
posture, required more accurate and reliable navigation
systems. Thenceforth, technology evolved rapidly: the
Navstar-GPS program was set in 1973, and in 1974, the
first atomic clock was put into orbit. Moving forward,
the first experimental Block-I GPS satellite was launched
in 1978.In 1983, Soviet jet interceptors shot down a Korean Air
civilian airliner carrying 269 passengers that had mistak-
enly entered Soviet airspace. Because the crew’s access to
better navigational tools might have prevented the disas-
ter, U.S. President Ronald Reagan issued a directive gua-
ranteeing that GPS signals would be available at no charge
for civilian uses when the system became operational. The
commercial market has grown steadily ever since. Second-generation GPS satellites were launched beginning in
1989, and the system’s full operational capability (FOC)
was declared in April 1995. Since initially the highest
quality signal was reserved for military use, the signal
available for civilian use was intentionally degraded [selec-
tive availability (SA)]. This changed with the U.S.
President Bill Clinton ordering SA turned off at midnight
May 1, 2000. This improved the potential precision ofcivilian GPS receivers from 300 to about 20 m.
In parallel, the former Soviet Union had begun the
development of the GLONASS system in 1976, also with
military endeavors, completing its satellite constellation in
1995. Following completion, the system rapidly fell into
disrepair with the collapse of the Russian economy. Begin-
ning in 2001, Russia committed itself to restoring the
system, and as of February 2011 it has been practicallyrestoredV22 satellites are operational. GLONASS pro-
vides two types of navigation signals: standard precision
(SP) navigation signal and high precision (HP) navigation
signal. SP positioning and timing services are available to
all GLONASS civil users on a continuous, worldwide basis.
Actually, on May 18, 2007, Russian President Vladimir
Putin signed a decree reiterating the offer to provide free
access to GLONASS civil signals.European’s Galileo program is the first guaranteed
global positioning service under civilian control. The first
stage of the Galileo program was agreed upon officially on
May 26, 2003 by the European Union and the European
Space Agency. By the end of 2013, it will have an initial
constellation of 16 satellites: 4 in-orbit validation (IOV)
and 12 FOC satellites. The European Commission an-
nounced that three of the five services offered by thesystem will be provided in early 2014: the open service
(basic signal provided free of charge), the public regulated
service (two encrypted signals with controlled access for
specific users like governmental bodies, agencies, and
organizations involved with defense, internal security, law
enforcement, and critical transport, providing position and
timing to specific users requiring a high continuity of
service), and the search and rescue service (contributing tothe international COSPAS-SARSAT cooperative system for
humanitarian search and rescue activities). The safety-of-
life service (an enhanced signal including an integrity
function that will warn the user within a few seconds in
case of a malfunction, intended for the safety-critical
transport community, e.g., aviation) and the commercial
service (combination of two encrypted signals for higher
data throughput rate and higher accuracy authenticateddata) will be tested in 2014 and will be provided as the
system reaches FOC with 30 satellites.
China is also deploying a GNSS named COMPASS. As
of February 2011, there were seven launched satellites, five
more will be launched before the end of 2012, and it was
announced that the FOC (comprising up to 35 satellites)
would be put in place by 2020.
These new deployments and system modernizationsdepict an unforeseen and close forthcoming scenario with
a plurality of systems, satellite constellations, frequency
bands, and new signal structures ready to be exploited.
From a situation in which GPS has been the mainVif not
the soleVplayer in satellite navigation, the rebirth of
GLONASS, the advent of Galileo and COMPASS, the in-
creasing regional system alternatives (QZSS or IRNSS),
and the worldwide maturity of augmentation systems(WAAS, EGNOS, MSAS, GAGAN) make the design of
future GNSS receivers more complex, since the number of
visible satellites is expected to increase from the current
10–13 values to 30–40. Multiconstellation/multifrequency
GNSS receivers, complemented with regional information,
promise dramatically improved positioning solutions and
enhanced integrity, at the expense of posing a number of
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1883
challenges to the navigation engineering community.Topics such as accuracy, precision, robustness, reliability,
interoperability with other systems, shorter time to first
fix, satellite selection, and coverage shall be tacked from
novel points of view, enabling unexplored business models
and applications that only imagination can bound.
This paper reviews imminently or already available
civil signals for satellite-based navigation in Section II,
with special emphasis on the benefits of new signals andcombining methods. Augmentation systems, both satellite
and ground based, are discussed in Section IV. Architec-
tural trends for GNSS receivers are inspected in Section V,
including advanced signal processing algorithms for signal
acquisition and tracking. Section VI shows how other
sources of information such as inertial sensors can be
integrated with the navigation system and interoperate
with it. Section VII takes a glimpse into the mass-marketperspective at the light of location-based services (LBSs)
and intelligent transport systems (ITSs). Finally,
Section VIII concludes the paper and draws topics for
further research.
II . GLOBAL NAVIGATION SATELLITESYSTEMS AND SIGNALS
GNSS space vehicles broadcast a low-rate navigation mes-
sage that modulates continuous repetitions of pseudoran-
dom spreading codes, which in turn are modulating a
carrier signal allocated in the L-band. The navigation mes-
sage, after proper demodulation, contains among other
information the so-called ephemeris, a set of parameters
that allow the computation of the satellite position at any
time. These positions, along with the corresponding dis-tance estimations, allow the receiver to compute its own
position and time, as we will see hereafter.
The distance between the receiver and a given satellite
can be computed by
�i ¼ c tRxi � tTx
i
� �(1)
where c ¼ 299 792 458 m/s is the speed of light, tRxi is the
receiving time in the receiver’s clock, and tTxi is the time of
transmission for a given satellite i. Receiver clocks are
inexpensive and not perfectly in sync with the satellite
clock, and thus this time deviation is another variable to be
estimated. The clocks on all of the satellites belonging to
the same system s, where s ¼ f GPS, Galileo, GLONASS,. . .g, are in sync with each other, so the receiver’s clock
will be out of sync with all satellites belonging to the same
constellation by the same amount �tðsÞ. In GNSS, the term
pseudorange is used to identify a range affected by a bias,
directly related to the bias between the receiver and satel-
lite clocks. There are other factors of error: since propaga-
tion at speed c is only possible in the vacuum, atmospheric
status affects the propagation speed of electromagneticwaves modifying the propagation time and thus the dis-
tance estimation. For instance, the ionosphere is a plasma-
tic medium that causes a slowdown in the group velocity
and a speedup of the phase velocity, having an impact in
code and phase delays and, thus, impeding precise naviga-
tion when its effects are not mitigated. Actually, errors can
be on the order of tens of meters in geomagnetic storm
episodes [3]. Ionospheric delay is a well-defined functionproportional to the inverse square of frequency and the
total electron content (TEC)Vwhich has diurnal, seasonal
and long-term variationsValong the signal path. Ac-
tually, this frequency dependence allows a multiband
receiver to remove the ionospheric effect by proper com-
binations of observations in different bands. As we will see
in Section IV, there are complementary systems that
inform purposely equipped receivers about the TEC statusin order to alleviate the impact of those effects, and thus
improve precision in the final navigation solution. This can
also be done by differential systems, where a network of
well-positioned, ground-fixed receivers measure the TEC
status and broadcast the corrections to the surrounding
rover receivers.
For each in-view satellite i of system s, we can write
�i ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixTx
i � xð Þ2þ yTxi � yð Þ2þ zTx
i � zð Þ2q
þ c�tðsÞ þ �e
(2)
where ðxTxi ; yTx
i ; zTxi Þ is the satellite’s position (known
from the navigation message), ðx; y; zÞ is the receiver’sposition, and �e gathers other sources of error. Since the
receiver needs to estimate its own 3-D position (three
spatial unknowns) and its clock deviation with respect to
the satellites’ time basis, at least 3þ Ns satellites must be
seen by the receiver at the same time, where Ns is the
number of different navigation systems available (in-view)
at a given time. Each received satellite signal, once syn-
chronized and demodulated at the receiver, defines oneequation such as the one defined in (2), forming a set of
nonlinear equations that can be solved algebraically by
means of the Bancroft algorithm [4] or numerically,
resorting to multidimensional Newton–Raphson and
weighted least square methods [5]. When a priori infor-
mation is added we resort to Bayesian estimation, a
problem that can be solved recursively by a Kalman filter
or any of its variants. The problem can be further expandedby adding other unknowns (for instance, parameters of
ionospheric and tropospheric models), sources of informa-
tion from other systems, mapping information, and even
motion models of the receiver. In the design of multi-
constellation GNSS receivers, the vector of unknowns can
also include the receiver clock offset with respect to each
system in order to take advantage of a higher number of
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1884 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
in-view satellites and using them jointly in the navigationsolution, therefore increasing accuracy.
The analytic representation of a signal received from a
GNSS satellite can be generically expressed as
rðtÞ ¼ �ðtÞsT t� �ðtÞð Þe�j2�fdðtÞej2�fct þ nðtÞ (3)
where �ðtÞ is the amplitude, sTðtÞ is the complex baseband
transmitted signal, �ðtÞ is the time-varying delay, fdðtÞ ¼fc�ðtÞ is the Doppler shift, fc is the carrier frequency, andnðtÞ is a noise term. These signals arrive to the Earth’s
surface at extremely low power (e.g., �158.5 dBW for GPS
L1 C/A-code, �157 dBW for Galileo E1), well below the
noise floor. In order to estimate its distances to satellites,
the receiver must correlate time-aligned replicas of the
corresponding pseudorandom code with the incoming
signal, in a process called despreading that provides pro-
cessing gain only to the signal of interest. After a coarseand fine estimation stages of the synchronization param-
eters (usually known as acquisition and tracking, respec-
tively), signal processing output is in the form of
observables: 1) the pseudorange (code) measurement,
equivalent to the difference of the time of reception
(expressed in the time frame of the receiver) and the time
of transmission (expressed in the time frame of the satel-
lite) of a distinct satellite signal; and optionally 2) thecarrier-phase measurement, actually being a measurement
on the beat frequency between the received carrier of the
satellite signal and a receiver-generated reference fre-
quency. Carrier phase measurements are ambiguous, in
the sense that the integer number of carrier wavelengths
between a satellite and the receiver’s antenna is
unknown. Techniques such as Least square AMBiguity
Decorrelation Approach (LAMBDA) [6] or Multi CarrierAmbiguity Resolution (MCAR) [7] can be applied to
resolve such ambiguity and provide an accurate estima-
tion of the distance between the satellite and the re-
ceiver. Then, depending on the required accuracy, the
navigation solution can range from pseudorange-only,
computationally low demanding, and limited accuracy
least squares methods to sophisticated combinations of
code and phase observables at different frequencies forhigh demanding applications such as surveying, geodesy,
and geophysics.
Next sections provide brief descriptions of the space
segment of different GNSSs and their broadcast signal
structures accessible by civilians.
A. Global Positioning System (GPS)The Navstar GPS [8] is a space-based radio-navigation
system owned by the United States Government (USG)
and operated by the United States Air Force (USAF). GPS
provides positioning and timing services to military and
civilian users on a continuous, worldwide basis. Two GPS
services are provided: the precise positioning service(PPS), available primarily to the military of the United
States and its allies, and the standard positioning service
(SPS) open to civilian users.
• GPS L1. Defined in [9], this band is centered at
fGPS L1 ¼ 1575.42 MHz. The complex baseband
transmitted signal can be written as
sðGPS L1ÞT ðtÞ ¼ eL1IðtÞ þ jeL1QðtÞ (4)
with
eL1IðtÞ ¼X1
l¼�1DNAV ½l�204600
� �
� CPðYÞ jljLPðYÞ
h ip t� lTc;PðYÞ� �
(5)
eL1QðtÞ ¼X1
l¼�1DNAV ½l�20460
� �
� CC=A jlj1023
� �pðt� lTc;C=AÞ (6)
where � is the exclusive-or operation (modulo-2
addition), jljL means l modulo L, ½l�L means the
integer part of l=L, DNAV is the GPS navigation
message bit sequence, transmitted at 50 b/s,
Tc;PðYÞ ¼ ð1=10:23Þ �s, Tc;C=A ¼ ð1=1:023Þ �s,
LPðYÞ ¼ 6:1871 � 1012, and pðtÞ is a rectangularpulse of a chip-period duration centered at t ¼ 0
and filtered at the transmitter. According to the
chip rate, the binary phase-shift keying modula-
tions in (5) and (6) are denoted as BPSK(10) and
BPSK(1), respectively. The precision P codes
(named Y codes whenever the antispoofing mode
is activated, encrypting the code and thus denying
non-U.S. military users) are sequences of sevendays in length. Regarding the modernization plans
for GPS, it is worthwhile to mention that there is a
new civilian-use signal planned, called L1C and
defined in [10], to be broadcast on the same L1
frequency that currently contains the C=A signal.
The L1C will be available with first Block III
launch, currently scheduled for 2013. The imple-
mentation will provide C=A code to ensurebackward compatibility.
• GPS L2C. Defined in [9], it is only available on
Block IIR-M and subsequent satellite blocks.
Centered at fGPS L2 ¼ 1227.60 MHz, the signal
structure is the same as in (4), with the precision
code in the in-phase component, just as in (5) but
with an optional presence of the navigation
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1885
message DNAV. For the quadrature-phase compo-nent, three options are defined
eL2CQðtÞ ¼X1
l¼�1DCNAV ½l�10230
� �
�
CCL jljLCL
h ip1=2ðt� lTc;L2CÞ þ CCM jljLCM
h ip1=2
� t� lþ 3
4
� �Tc;L2C
� �!(7)
eL2CQðtÞ ¼X1
l¼�1DNAV ½l�20460
� �� CC=A jlj1023
� �pðt� lTc;C=AÞ
(8)
or
eL2CQðtÞ ¼X1
l¼�1CC=A jlj1023
� �pðt� lTc;C=AÞ (9)
where Tc;L2C ¼ (1/511.5) ms and p1=2ðtÞ is a rec-
tangular pulse of half chip-period duration, thus
time-multiplexing both codes. The civilian long code
CCL is LCL ¼767 250 chips long, repeating every1.5 s, while the civilian moderate code CCM is
LCL ¼ 10 230 chips long and its repeats every 20 ms.
The CNAV data is an upgraded version of the
original NAV navigation message, containing higher
precision representation and nominally more accu-
rate data than the NAV data. It is transmitted at
25 b/s with forward error correction (FEC) encod-
ing, resulting in 50 symbols per second (sps).• GPS L5. The GPS L5 link, defined in [11], is only
available in Block IIF (first satellite launched on
May, 2010) and subsequent satellite blocks.
Centered at fGPS L5 ¼ 1176:45 MHz, this signal
in space can be written as:
sTðtÞðGPS L5Þ ¼ eL5IðtÞ þ jeL5QðtÞ (10)
eL5IðtÞ ¼Xþ1
m¼�1Cnh10
jmj10
� �� DCNAV ½m�10
� �
�X102300
l¼1
CL5I jlj10230
� �pðt� mTc;nh � lTc;L5Þ
(11)
eL5QðtÞ ¼Xþ1
m¼�1Cnh20
jmj20
� ��X102300
l¼1
CL5Q jlj10230
� �� pðt� mTc;nh � lTc;L5Þ (12)
where Tc;nh ¼ 1 ms and Tc;L5 ¼ (1/10.23) �s, thusdefining a BPSK(10) modulation. Both L5I and
L5Q contain synchronization sequences.
B. GLONASSThe nominal baseline constellation of the Russian Fed-
eration’s Global Navigation Satellite System (GLONASS)
comprises 24 GLONASS-M satellites that are uniformly
deployed in three roughly circular orbital planes at an
inclination of 64.8� to the equator. The altitude of theorbit is 19 100 km. The orbit period of each satellite is 11 h,
15 min, and 45 s. The orbital planes are separated by 120�
right ascension of the ascending node. Eight satellites are
equally spaced in each plane with 45� argument of latitude.
Moreover, the orbital planes have an argument of latitude
displacement of 15� relative to each other.
GLONASS civil signal-in-space is defined in [12]. This
system makes use of a frequency-division multiple-access(FDMA) signal structure, transmitting in two bands:
fðkÞGLO L1 ¼ 1602þ k � 0:5625 MHz and f
ðkÞGLO L2 ¼ 1246 þ
k � 0:4375 MHz, where k 2 f�7;�6; . . . ; 5; 6g is the
channel number. Satellites in opposite points of an orbit
plane transmit signals on equal frequencies, as these satel-
lites will never be in view simultaneously by a ground-
based user.
• GLONASS L1. Two kinds of signals are trans-mitted: an SP and an obfuscated HP signal. The
complex baseband transmitted signal can be
written as
sTðtÞðGLO L1Þ ¼ eL1IðtÞ þ jeL1QðtÞ (13)
with BPSK(5) and BPSK(0.5) modulations
eL1IðtÞ ¼X1
l¼�1DGNAV ½l�102200
� �� CHP jljLHP
� �pðt� lTc;HPÞ
(14)
eL1QðtÞ ¼X1
l¼�1DGNAV ½l�10220
� �� CSP jlj511
� �pðt� lTc;SPÞ
(15)
where Tc;HP ¼ (1/5.11) �s, Tc;SP ¼ (1/0.511) �s,
and LHP ¼ 3:3554 � 107. The navigation message
DGNAV is transmitted at 50 b/s. Details of its con-tent and structure, as well as the generation of the
CSP code, can be found in [12]. The usage of the HP
signal should be agreed with the Russian Federa-
tion Defense Ministry, and no more details have
been disclosed.
• GLONASS L2. Beginning with the second gener-
ation of satellites, called GLONASS-M and first
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1886 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
launched in 2001, a second civil signal is availableusing the same SP code as the one in the L1 band.
The use of FDMA techniques, in which the same code is
used to broadcast navigation signals on different fre-
quencies, and the placement of civil GLONASS transmis-
sions on frequencies close to 1600 MHz, well above the
GPS L1 band, have complicated the design of combined
GLONASS/GPS receivers, particularly low-cost equipment
for mass-market applications. Future plans of moderni-zation are intended to increase compatibility and inter-
operability with other GNSS, and include the addition of a
code-division multiple-access (CDMA) structure, and
possibly binary offset carrier (BOC) modulation, begin-
ning with the third civil signal in the L3 band (1197.648–
1212.255 MHz). Russia is implementing the new signals
on the next-generation GLONASS-K satellites, with a
first prototype successfully launched into orbit onFebruary 26, 2011.
C. GalileoThe nominal Galileo constellation comprises a total of
27 operational satellites (plus three active spares), which
are evenly distributed among three orbital planes inclined
at 56� relative to the equator. There are nine operational
satellites per orbital plane, occupying evenly distributed
orbital slots. Three additional spare satellites (one per or-
bital plane) complement the nominal constellation confi-
guration. The Galileo satellites are placed in quasi-circularEarth orbits with a nominal semi-major axis of about
30 000 km and an approximate revolution period of 14 h.
The control segment full infrastructure will be composed
of 30–40 sensor stations, three control centers, nine mis-
sion uplink stations, and five TT&C stations.
Galileo’s open service is defined in [13], where the
following signal structures are specified.
• Galileo E1. This band, centered at fGal E1 ¼1575.420 MHz and with a reference bandwidth of
24.5520 MHz, uses the so-called composite binary
offset carrier CBOC(6,1,1/11) modulation, defined
in baseband as
sðGal E1ÞT ðtÞ ¼ 1ffiffiffi
2p eE1BðtÞ �scAðtÞ þ �scBðtÞð Þ þð
� eE1CðtÞ �scAðtÞ � �scBðtÞð ÞÞ (16)
where the subcarriers scðtÞ are defined as
scAðtÞ ¼ sign sinð2�fs;E1AtÞ� �
(17)
scBðtÞ ¼ sign sinð2�fs;E1BtÞ� �
(18)
and fs;E1A ¼ 1.023 MHz, fs;E1B ¼ 6.138 MHz are the
subcarrier rates, � ¼ffiffiffiffiffiffiffiffiffiffiffi10=11
p, and � ¼
ffiffiffiffiffiffiffiffiffi1=11
p.
Channel B contains the I/NAV type of navigationmessage DI=NAV, intended for safety-of-life (SoL)
services
eE1BðtÞ ¼Xþ1
l¼�1DI=NAV ½l�4092
� �� CE1B jlj4092
� �pðt� lTc;E1BÞ:
(19)
In case of channel C, it is a pilot (dataless) channel
with a secondary code, forming a tiered code
eE1CðtÞ ¼Xþ1
m¼�1CE1Cs jmj25
� ��X4092
l¼1
CE1Cp½l�
� pðt� mTc;E1Cs � lTc;E1CpÞ (20)
with Tc;E1B ¼ Tc;E1Cp ¼ (1/1.023) �s and Tc;E1Cs ¼4 ms. The CE1B and CE1Cp primary codes are
pseudorandom memory code sequences defined
in [13, Annex C.7 and C.8]. The binary sequence of
the secondary code CE1Cs is 0011100000001010110
110010. This band also contains another compo-
nent, Galileo E1A, intended for the public regu-
lated service (PRS). The PRS spreading codes andthe structure of the navigation message have not
been made public.
• Galileo E6. Intended for the commercial service
and centered at fGal E6 ¼ 1278.750 MHz, this band
provides pilot and data components
sðGal E6ÞT ðtÞ ¼ 1ffiffiffi
2p eE6BðtÞ � eE6CðtÞð Þ (21)
eE6BðtÞ ¼Xþ1
m¼�1DC=NAV ½l�5115
� �E6BjljLE6B
h i� pðt� lTc;E6Þ
(22)
eE6CðtÞ ¼Xþ1
m¼�1CE6Cs jmj100
� ��XLE6C
l¼1
CE6Cp½l�
� pðt� mTc;E6s � lTc;E6pÞ (23)
where DC=NAV is the C/NAV navigation data
stream, which is modulated with the encryptedranging code CE6B with chip period Tc;E6 ¼(1/5.115) �s, thus being a BPSK(5) modulation.
Codes CE6B and primary codes CE6Cs and their
respective lengths LE6B and LE6C have not been
published. The secondary codes for the pilot
component CE6Cs are available in [13]. The receiver
reference bandwidth for this signal is 40.920 MHz.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1887
This band also contains another component,Galileo E6A, intended for PRS.
• Galileo E5. Centered at fGal E5 ¼ 1191.795 MHz
and with a total bandwidth of 51.150 MHz, its sig-
nal structure deserves some analysis. The AltBOC
modulation can be generically expressed as
sAltBOCðtÞ ¼ x1ðtÞv�ðtÞ þ x2ðtÞvðtÞ (24)
w h e r e vðtÞ ¼ ð1=ffiffiffi2pÞðsignðcosð2�fstÞÞ þ
jsignðsinð2�fstÞÞÞ is the single sideband subcar-
rier, fs is the subcarrier frequency, ð�Þ� stands for
the conjugate operation, and x1ðtÞ and x2ðtÞ areQPSK signals. The resulting waveform does not
exhibit constant envelope. In case of Galileo, the
need for high efficiency of the satellites’ onboard
high-power amplifier (HPA) has pushed a modi-
fication on the signal in order to make it envelope
constant and thus use the HPA at saturation. This
can be done by adding some intermodulation
products to (24), coming up with the followingdefinition:
sðGal E5ÞT ðtÞ ¼ eE5aðtÞssc�s ðtÞ þ eE5bðtÞsscsðtÞ
þ �eE5aðtÞssc�pðtÞ þ �eE5bðtÞsscpðtÞ (25)
where the single and product sideband signal sub-carriers are
sscsðtÞ ¼ scsðtÞ þ jscs t� Ts
4
� �(26)
sscpðtÞ ¼ scpðtÞ þ jscp t� Ts
4
� �(27)
and
eE5aðtÞ ¼ eE5aIðtÞ þ jeE5aQðtÞ (28)
eE5bðtÞ ¼ eE5bIðtÞ þ jeE5bQðtÞ (29)
�eE5aðtÞ ¼ �eE5aIðtÞ þ j�eE5aQðtÞ (30)
�eE5bðtÞ ¼ �eE5bIðtÞ þ j�eE5bQðtÞ (31)
�eE5aIðtÞ ¼ eE5aQðtÞeE5bIðtÞeE5bQðtÞ (32)
�eE5aQðtÞ ¼ eE5aIðtÞeE5bIðtÞeE5bQðtÞ (33)
�eE5bIðtÞ ¼ eE5bQðtÞeE5aIðtÞeE5aQðtÞ (34)
�eE5bQðtÞ ¼ eE5bIðtÞeE5aIðtÞeE5aQðtÞ: (35)
The signal components are defined as
eE5aIðtÞ ¼Xþ1
m¼�1CE5aIs jmj20
� ��X10230
l¼1
CE5aIp½l�
� DF=NAV ½l�204600
� �pðt� mTc;E5s � lTc;E5pÞ (36)
eE5aQðtÞ ¼Xþ1
m¼�1CE5aQs jmj100
� ��X10230
l¼1
CE5aQp½l�
� pðt� mTc;E5s � lTc;E5pÞ (37)
eE5bIðtÞ ¼Xþ1
m¼�1CE5bIs jmj4
� ��X10230
l¼1
CE5aIp½l�
� DI=NAV ½l�40920
� �pðt� mTc;E5s � lTc;E5pÞ (38)
eE5bQðtÞ ¼Xþ1
m¼�1CE5bQs jmj100
� ��X10230
l¼1
CE5bQp½l�
� pðt� mTc;E5s � lTc;E5pÞ (39)
where Tc;E5s ¼ 1 ms and Tc;E5p ¼ (1/10.23) �s.
Channel A contains the F/NAV type of navigation
message DF=NAV, intended for the open service.
The I/NAV message structures for the E5bI andE1B signals use the same page layout. Only page
sequencing is different, with page swapping be-
tween both components in order to allow a fast
reception of data by a dual frequency receiver. The
single subcarrier scsðtÞ and the product subcarrier
scpðtÞ of (26) and (27) are defined as
scsðtÞ ¼ffiffiffi2p
4sign cos 2�fst�
�
4
þ 1
2sign cosð2�fstÞð Þ
þffiffiffi2p
4sign cos 2�fstþ
�
4
(40)
scpðtÞ ¼ �ffiffiffi2p
4sign cos 2�fst�
�
4
þ 1
2sign cosð2�fstÞð Þ
þ �ffiffiffi2p
4sign cos 2�fstþ
�
4
(41)
with a subcarrier frequency of fs ¼ 15.345 MHz,
thus defining an AltBOC(15,10) modulation. Plot-
ting the power spectrum of the carriers in (25) (see
Fig. 1), we can see that the QPSK(10) signal eE5aðtÞdefined in (25) is shifted to fGal E5a¼: fGal E5 � fs ¼1176.450 MHz, while eE5bðtÞ is shifted to
fGal E5b¼:
fGal E5 þ fs ¼ 1207:140 MHz . Thus, we
can bandpass filter around fGal E5a and get a goodapproximation of a QPSK(10) signal, with very low
energy components of eE5bðtÞ, �eE5aðtÞ, and �eE5bðtÞ:
sðGal E5aÞT ðtÞ ’ eE5aIðtÞ þ jeE5aQðtÞ: (42)
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1888 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
The same applies to eE5bðtÞ, allowing an indepen-
dent reception of two QPSK(10) signals and thus
requiring considerably less bandwidth than the
processing of the whole E5 band. A hardware
architecture proposal for an AltBOC(15,10) receiv-
er is found in [14].
D. COMPASSThe Chinese COMPASS navigation satellite system will
consist of five geostationary satellites and 30 nongeosta-
tionary satellites. The geostationary satellites will be lo-
cated at 58.75� E, 80� E, 110.5� E, 140� E, and 160� E.
Nongeostationary satellites will be in medium-Earth orbit
(MEO) and inclined geosynchronous orbit. After the first
satellite (located at 140� E) was launched on October 31,2000, a second satellite (located at 80� E) and a third
satellite (located at 110.5� E) were launched on December
21, 2000 and May 25, 2003, respectively. The first MEO
satellite, named COMPASS-M1, was launched on April 14,
2007, and its signals have been unraveled by independent
research [15]–[17]. The first geostationary satellite
COMPASS-G2 was launched on April 15, 2009, and the
second one G3 on June 2, 2010. Global coverage is plannedby 2020. The ground segment will consist of one master
control station, two upload stations, and 30 monitor sta-
tions. This system will provide three frequency bands: B1,
centered at 1561.098 MHz and carrying a QPSK(2) signal
with a 4-MHz bandwidth; B2, centered at 1207.14 MHz,
with one BPSK(2) and one BPSK(10) with a 24-MHz
bandwidth; and B3, centered at 1268 and carrying a
QPSK(10) with the same bandwidth as B2. To the authors’knowledge, no official interface control document has
been made public as of February 2011.
E. Benefits of New Signals and Combining MethodsPotential benefits of new signals and their combina-
tions are expected to be high in terms of accuracy and
reliability. From a signal processing perspective, power
spectrum shapes of new signals allow for improved perfor-
mance in time-delay estimation. The Cramer–Rao lower
bound (CRLB) is the theoretical limit of variance that any
unbiased estimator can achieve. In case of the time-delayestimation, its minimum variance is defined by [18]
�2error
1
2E
N0
� �ð2�Þ2
Z 1�1
f 2SrðfÞ df
(43)
where E=N0 is the received-signal-energy-to-noise-power-
spectral-density ratio, f is the frequency in hertz, and SrðfÞis the normalized power spectral density of the code
modulation. Therefore, the integral represents the second
moment of SrðfÞ. The square root of this second moment is
commonly referred to as RMS or Gabor bandwidth [19],and it is a fundamental parameter in code modulation de-
sign in order to theoretically achieve a smaller time-delay
error
BG ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiZ 1�1
f 2SrðfÞ df
s: (44)
Inspecting (44), it can be observed that the Gabor
bandwidth can be increased by using modulations whose
power spectrum concentrates a greater percentage of their
power farther from the signal center frequency, because of
the weighting term f 2. This approach has been followed inthe definition of the modernized and forthcoming GNSS
signals with the introduction of the BOC modulation,
where a square carrier of frequency fs splits the main lobe
of the code spectrum into two lobes centered at fs from
the central frequency, thus allocating more power at
higher frequencies and potentially improving synchroni-
zation performance.
From a multiband receiver perspective, more fre-quency bands means more possibilities to resolve the un-
known cycle ambiguities of the double-differenced carrier
phase data to integers, a key aspect for rapid and very
precise (centimeter-level) GNSS positioning. Multiple
bands also allow linear combinations of observables that
can be used to eliminate or mitigate individual sources
of error (e.g., the ionospheric effect can be removed by
Fig. 1. Power spectrum of single and product sideband subcarriers
signals in (25), normalized to the power of ssc�s ðtÞ at fGal E5a. The
modified AltBOC modulation can be well approximated by two QPSK
signals 2fs apart, with negligible contribution of the crossed terms
around its center frequency.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1889
exploiting its frequency dependence), and to alleviateexcessive computational burden [20].
From a multiconstellation receiver perspective, in-
creased coverage and reliability are the expected outcomes
as more satellites will be available. Novel approaches to
positioning such as direct position estimation [21] could
deeply combine signals and further increase positioning
performance in hostile (radioelectrically speaking) scenar-
ios, as shown in [22].
III . REGIONAL NAVIGATION SATELLITESYSTEMS (RNSS)
Regional systems provide additional signals from satellites
operating over a given geographical area that are compa-
tible with one or more GNSS systems. Currently there are
two main RNSSs.
A. Quasi-Zenith Satellite System (QZSS)The QZSS is designed to work in conjunction with, and
to enhance, the civil services of other GNSSs in Japan,
particularly in urban environments where buildings block
visibility of a great portion of the sky. The planned QZSS
constellation consists of three satellites that would be
placed in highly elliptical periodic orbits (HEO), atgeosynchronous altitude (around 35 786 km) inclined
43� to the equatorial plane. The three orbital planes draw
the same eight-shaped ground track, with the northern
loop covering a much smaller geographical area (mainly
the central part of Japan) than the southern one (covering
Australia and New Guinea), with a central line at 135� E in
longitude. These orbits allow the satellites to dwell for
more than 12 h/day with an elevation above 70�, meaningthat they appear almost overhead most of the time, and
giving rise to the term Bquasi-zenith[ for which the system
is named. The first satellite was launched on September 11,
2010.
QZSS satellites will transmit six positioning signals:
L1C/A, L1-SAIF, and L1C signals centered at 1575.42 MHz
(common with GPS L1 and Galileo E1), the L2C signal
centered at 1227.60 MHz (common with GPS L2C), theLEX signal at 1278.75 MHz (common with Galileo E6),
and the L5 signal at 1176.45 MHz (common with GPS L5).
L1C/A, L1C, L2C, and L5 will be positioning availabilityenhancement signals, because they will complement exist-
ing GNSS. L1-SAIF and LEX will be positioning performanceenhancement signals, transmitting differential data of
existing GNSS and integrity data concerning GNSS signals
as determined by QZSS. Although the interface specifica-tion document is still in draft [23], L1 and L2 signals
should be defined with a bandwidth of 24 MHz, L5 signals
with 24.9 MHz, and LEX with 42 MHz. Signals in L1, L2,
and L5 will be very similar to their counterparts in GPS,
making use of pseudorandom codes defined in the GPS’s
interface specification documents [9]–[11] but not used by
any GPS satellite.
B. Indian Regional NavigationSatellite System (IRNSS)
The IRNSS, developed by the Indian Space Research
Organization, will offer a standard positioning service
using BPSK(1) modulation as well as a restricted/autho-
rized service employing a BOC(5,2) modulation. Both of
these services will be provided at two frequencies in the L5
and S bands. The space segment will comprehend seven
satellites: three of them in a GEO orbit at 34�, 83�, and132� E, while the other four will have a geosynchronous
orbit at 29� inclination, with longitude crossing at 55� and
111� E [24]. The first satellite is scheduled to launch in the
first half of 2012 [25]. The system is intended to provide an
all-weather absolute position accuracy ð2�Þ of better than
20 m throughout India and within a region extending
approximately 1500 km around it.
IV. AUGMENTATION SYSTEMS
GNSS receivers’ performance can be augmented by means
of complementary systems. Accuracy (difference between
the estimated position and the true position), integrity (a
measure of trust placed in the correctness of the infor-
mation provided by the navigation system, and its ability to
provide timely warnings), availability (the probability thatthe navigation and fault detection functions are opera-
tional), continuity (the ability to provide the navigation
function over time), and time to first fix can be drama-
tically enhanced when additional information external to
the receiver is provided. For instance, the receiver can be
precisely informed about the ionosphere status (i.e., the
electron content), allowing the removal of the bias it
provokes in the observables, and thus improving the accu-racy of the final navigation solution. Another example is
integrity, a key factor in safety-critical applications. Users
may determine their integrity by receiver autonomous
algorithms (RAIM) [26], or by using external integrity data
sources. Galileo foresees the provision of integrity data
within the navigation message, although this service will
not likely be implemented in the first FOC phase. Usually,
external sources (the so-called augmentation systems)consist of a set of fixed, accurately surveyed receivers
forming a network of reference (or control, or fiducial)
stations. The observations taken by this reference network
are broadcast to the rover GNSS receiver, which processes
them simultaneously with its own observations, removing
common sources of error, and detecting inconsistencies in
the navigation solution based on the exploitation of redun-
dant GNSS measurements. Hereafter, we provide descrip-tions of such augmentation systems, classified according to
the way how this external information is sent to the
receiver.
A. Satellite-Based Augmentation System (SBAS)An SBAS supports wide area or regional augmentation
through the use of additional satellite-broadcast messages.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1890 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
Such systems are commonly composed of multiple groundstations, located at accurately surveyed points. The ground
stations take measurements of one or more of the GNSS
satellites, the satellite signals, or other environmental fac-
tors which may affect the signal received by the users.
Using these measurements, information messages are
created and sent to geostationary satellites for broadcast
to the end users. While SBAS implementations may vary
widely, rules from the International Civil Aviation Organi-zation (ICAO) establish that an SBAS must transmit a
specific message format and frequency, as defined in [27]
for civil aviation use. Actual implementations are as
follows.
1) Wide-Area Augmentation System (WAAS): The WAAS
[28], operated by the U.S. Federal Aviation Administra-
tion, has declared to offer continuous service for nonsafetyapplications since August 2000, and it was commissioned
for safety-of-life services in July 2003. The system is
composed of two geostationary satellites located at 133� W
and 107� W, broadcasting GPS-like ranging signals at both
GPS L1 and L5 carrier frequencies [29], and a ground
network with monitors throughout the United States,
Canada, and Mexico. A group of master stations collect
measurements from this reference stations network, buildthe SBAS message, and upload the message to the GEO
satellites. The WAAS improves accuracy in two funda-
mental ways: 1) it reduces range measurement error by
sending differential corrections for each GPS satellite,
reducing the pseudorange measurement error from around
30 m to approximately 1 or 2 m ð1�Þ, and 2) it improves
geometry by adding new ranging signals to the set of
available GPS measurements, since the phase of the WAAScode is synchronized to GPS time. The system also moni-
tors integrity, broadcasting error bounds for each moni-
tored satellite and updating this information within 6 s of
any significant change.
2) European Geostationary Navigation Overlay Service(EGNOS): The EGNOS [30] consists of three geostationary
satellites and a network of ground stations, operative since2009. It is intended to supplement the GPS, GLONASS,
and Galileo systems by reporting on the reliability and
accuracy of the signals. Specifications state that horizontal
position accuracy should be better than 7 m, but in practice
it is better than 2 m with an availability above 99%.
Document [31] details the general conditions relating to
the use of the EGNOS service, a technical description of
the signal-in-space, the reference receiver, environmentalconditions, the service performance achieved, and aspects
relating to service provision. EGNOS will offer also a
commercial service, a ground-based access to its data
through EGNOS data access service (EDAS), and a SoL
service [32]. Main applications of EGNOS are aviation,
precision agriculture, maritime, land transportation, and
time standard.
3) Others: Other SBAS are the Multifunctional SatelliteAugmentation System (MSAS), operated by Japan’s
Ministry of Land, Infrastructure and Transport, that was
commissioned for aviation use on September 27, 2007, the
GPS and GEO Augmented Navigation (GAGAN) planned
by India [33], and the Russian System for Differential
Correction and Monitoring (SDCM). Commercial systems
include StarFire [34] and OmniSTAR [35], offering sub-
decimetric accuracy. There also exists a proposal of aUniversal-SBAS (U-SBAS) standard [36] that carries addi-
tional channels (signals and messages) to cover the nonae-
ronautical specific SoL services, and also: high precision
positioning services, position velocity time (PVT), authen-
tication services, safety services, scientific application ser-
vices, high precision timing services, etc. U-SBAS is
designed to be fully interoperable with the current SBAS
standards and to allow significant performance and serviceimprovements in operational, scientific, and/or security
areas.
B. Ground-Based Augmentation Systems (GBAS)Systems that support augmentation through the use of
terrestrial radio messages are called GBASs. As with
SBASs, GBASs are commonly composed of one or more
accurately surveyed ground stations, which take measure-ments concerning the GNSS, and one or more radio trans-
mitters, which transmit the information directly to the end
user. These systems are usually devoted to air traffic man-
agement, specifically aircraft landing guidance. The net-
work of reference stations are localized around an airport,
supporting receivers within 20 km, and the differential
corrections and integrity information are broadcast over a
very high-frequency (VHF) data broadcast (VDB) signaltransmitted in the 108.0–117.975-MHz band [37].
1) Ground-Based Regional Augmentation System (GRAS):They support a large regional area. A network of fixed,
reference stations broadcast the difference between the
positions indicated by the satellite systems and their
known fixed positions. These stations broadcast the dif-
ference between the measured satellite pseudoranges andactual (internally computed) ones, so that receiver devices
may correct their pseudoranges by the same amount. The
correction signal is typically broadcast by an ultrahigh-
frequency (UHF) radio modem. Since satellite ephemeris
errors and atmospheric errors vary with space, accuracy
degrades with the distance to the reference stations.
Precision can be enhanced by exploiting carrier phase
measurements. Real-time kinematic (RTK) positioning hasbecome an industry standard procedure in surveying, lab
prototypes for research activities, machine control, and
other high-precision applications. RTK makes use of
carrier-phase and pseudorange measurements recorded
at a (usually) fixed reference location with known coordi-
nates and transmitted in real time to a user’s rover receiver
using a radio link of some kind. The rover processes the
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1891
double differences of observations to determine its coordi-nates with accuracy better than 10 cm. This method as-
sumes that the differential ionospheric delay between the
reference and the rover receiver is negligible, which works
well for baselines up to 10–20 km. The coverage can be
enlarged by a set of reference receivers, resorting to net-
work RTK (NRTK). This approach achieves real-time
subdecimeter error level positioning with distances up to
50–70 km.This concept has been extended by exploiting the full
geometry of the observations in a real-time ionospheric
model of the slant delay: a central processing facility (CPF)
combines new ionospheric tomography and traveling
ionospheric disturbance models with real-time undiffer-
enced processing of measurements from widely separated
permanent GNSS receivers. The CPF provides the rover
receivers with undifferenced accurate ionospheric correc-tions in real time, which are used to enhance the final
position estimation up to typical accuracies of around
10 cm of error, within a short convergence time [38]. This
approach, known as wide-area RTK (WARTK), allows for
permanent stations 500–900 km apart (thus implying a
dramatic reduction of reference stations with respect to
NRTK), and it is able to provide coverage at a continental
scale.
2) Assisted GNSS: Existing wireless infrastructure can be
used to broadcast information that would allow the re-
ceiver to perform better in terms of time to fix, sensitivity,
and accuracy, an approach known as assisted GNSS. For
instance, if navigation data are made available to the re-
ceiver via GPRS, WiFi or other, it can notably reduce its
time to first fix (TTFF) and improve receiver sensitivity.Furthermore, the delivery of precise orbit predictions can
also help to improve the receiver performance [39].
A-GNSS is becoming extremely popular in low-cost de-
vices, thanks to the wide current coverage of the cellular
communications networks and industrial standardization.
Cellular industry location standards first appeared in
the late 1990s, with the 3rd generation partnership (3GPP)
radio resource location services protocol (RRLP) technicalspecification 44.031 positioning protocol for GSM net-
works. Today, RRLP is the de facto standardized protocol to
carry GNSS assistance data to GNSS-enabled mobile de-
vices [40]. A major update began in 2007, when RRLP
release 7 added support for assisted-Galileo, and release
8 for the rest of the GNSS including the various SBAS. The
two releases provided native assistance data types such as
global Klobuchar and NeQuick models for the ionosphere[30], [41]. The same approach was also mapped to 3GPP TS
25.331 radio resource control (RRC) protocol, which
defines the positioning procedures and assistance data
delivery for UMTS terrestrial radio access (UTRA).
3GPP boosted location services in long-term evolution
(LTE) release 9, frozen in December 2009. According to
the LTE location architecture, the evolved serving mobile
location center (E-SMLC) is the server component incharge of positioning activities. The mobility management
entity (MME) gives the positioning request to E-SMLC,
which then controls the user equipment to be positioned
and, possibly, LTE base stations, to perform positioning.
The actual positioning and assistance protocol between E-
SMLC and the user equipment is called LTE positioning
protocol (LPP).
RRLP, RRC, and LPP are natively control-plane posi-tioning protocols, i.e., they use cellular signaling channels
as the transport mechanism for the assistance data and
position information. Since signaling channels are not de-
signed to transfer large amount of information, in 2003,
the open mobile alliance (OMA) started to work with se-
cure user plane location (SUPL) 1.0 protocol, which brings
the same location capabilities to user plane (and thus the
traffic channels) over IP networks as RRLP/RRC/LPP bringto control plane. SUPL 1.0 is already commercially de-
ployed, and SUPL 2.0 is now being deployed globally.
These protocols typically address richer GNSS features for
LBSs.
LTE release 9 introduced extension hooks in LPP mes-
sages, so that the bodies external to 3GPP could extend the
LPP feature set. OMA LPP extensions (LPPe), supported in
SUPL 3.0, build on top of the 3GPP LPP, reusing its pro-cedures and data types as far as possible. This ensures that
in the user-plane domain, which dominates in consumer
LBSs (LBS), vendors can utilize exactly the same protocol
as in the control plane. This reduces implementation,
testing, and deployments costs, and probably will make the
LPP/LPPe the de facto standardized positioning protocol in
the mobile domain [40].
Two different methodologies of assistance have beenstandardized, usually referred to as mobile-station (MS)-
based (UE-based) and MS-assisted (UE-assisted). In the
MS-based approach, the network operator provides the
GNSS-enabled mobile device with assistance data such as,
at a minimum, an approximate location coming from the
serving cell tower, an approximate time (accurate to a few
seconds), and a description of the satellite orbits and clock
errors (navigation model); the receiver uses that informa-tion to estimate the expected delays and Doppler shifts of
the visible satellites and proceeds to the acquisition of
satellite signals with a narrower search space, allowing a
dramatic reduction of the TTFF. Finally, location infor-
mation is sent back to the network in response to the
location request.
In the MS-assisted approach, assistance data consist in
the list of visible satellites, expected delays, and expectedDoppler shifts. Then, the GNSS receiver performs acqui-
sition and sends measurements (delay, Doppler frequency,
and signal power to noise ratio) back to a server in the
operator’s network, which computes subscriber’s position.
In general, MS-based methods are preferred over MS-
assisted methods since they are advantageous in terms
of position accuracy, allow the use of sophisticated
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1892 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
(Kalman-like) navigation filters, and are well suited tocontinuous navigation by delivering ephemeris data. How-
ever, the amount of data exchanged in the MS-based ap-
proach is significantly larger than that of the MS-assisted
one (currently about seven times), and is expected to
increase in short with the addition of new GNSS
satellites.
V. TECHNOLOGIES ANDARCHITECTURES FOR GNSS RECEIVERS
Global marketplace for GNSS receivers has been tradi-
tionally based on application specific integrated circuit
(ASIC) technology, an approach with high development
costs but extremely low cost per unit (thus ensuring re-
venues) and high performance. However, recent and
forthcoming changes in the space segment are pushingdevelopers to new approaches and targeting designs to
unforeseen levels of accuracy, reliability, and availability.
This requires more flexibility in the design and implemen-
tation processes, driving to more agile development tools.
Although ASIC technology remains pervasive for mass
market applications, other technologies such as field-
programmable gate array (FPGA) or software-defined
radio (SDR) running in digital signal processor (DSP),microprocessors, or even regular PCs are of great interest
for limited market but highly demanding applications such
as reference stations, geodesy and surveying, timing, ma-
chine control, and data gathering for scientific purposes
(study of ionosphere, troposphere, weather forecasting,
etc.).
A. ASIC TechnologyToday, established chipsets and original equipment
manufacturer (OEM) modules dominate the market, spe-
cially mass-market applications in the form of smart-
phones, digital cameras and camcoders, portable gaming
consoles, media players, and personal navigation devices.
Those chips are ASICs, nonstandard integrated circuits
that have been designed for a specific use or application.
Generally, an ASIC is undertaken for a product that willhave a large production run, since the cost of an ASIC
development is high and thus targeted to high volume
productions. In commercial receivers, ASICs are used to
handle the downconversion and digitizing, and to carry
out massively parallel correlation operations, while micro-
processors such as an ARM processor are utilized for
baseband signal processing. Current technology allows for
a higher level of integration: there are commercial solu-tions that integrate a whole GPS receiver (from the
antenna to the output in NMEA 0183 standard data format
that includes position, velocity, and time) in a single chip
[42]–[44].
The usual way to design ASIC is in the form of intel-
lectual property (IP) core of the designed circuits, which
completely defines the chip-level solution and allows a
mass production that reduces the overall fabrication costand time. There are a number of radio-frequency (RF) IPs
and baseband IPs commercially available. According to a
recent market study [45], Broadcom Corporation was the
leading supplier of GPS chips worldwide in March 2009.
SiRF Technology, Inc. and Texas Instruments, Inc. came in
second and third place in the GPS integrated circuit
manufacturer vendor list. Then followed Qualcomm Inc.,
STMicroelectronics, u-blox AG, Atheros Communications,Inc., Infineon Technologies AG, Atmel Corporation, and
MediaTek, Inc.
Since the vast majority of existing integrated circuits
are intended for the GPS L1 C/A signal, they make use of a
2-MHz RF bandwidth, which is enough for the BPSK
signal defined in (6) but prevent from the use of the
Galileo E1 signal defined in (16)–(20) due to the charac-
teristics of the CBOC split-spectrum nature, which re-quires at least 4-MHz (although 8 MHz is recommended;
see [46]) RF bandwidth for proper synchronization. New
signals will push new designs addressing both challenging
electrical requirements (wider bandwidths, more fre-
quency bands) and associated computational load. Even-
tually, the mass market will pose requirements in terms of
accuracy, coverage, and reliability that existing GPS-only
designs would not be able to meet, and therefore anevolution in GNSS chipset design is envisaged. The trend
seems to point towards an increasing integration of tech-
nologies (i.e., cellular [47], WiFi [48], RFID [49], UWB
[50], wireless sensor networks [51], or motion sensors
[52]), vertical integration and disappearance of indepen-
dent chipset manufacturers, lower cost, lower power
consumption, improved sensitivity, and smaller footprint.
As discussed in [53], the Galileo’s E1/E5a combinationcould be the best dual-frequency solution for a Galileo
mass-market receiver, since it overlaps exactly with GPS’
L1/L5 frequency bands and would allow multiband/
multiconstellation receivers at reasonable electrical and
computational requirements.
B. Software-Defined Radio ReceiversThe last decade has witnessed a rapid evolution of
GNSS software receivers. Since the first GPS standard
positioning service software receiver described in [54],
where the concept of bandpass sampling (or intentional
aliasing) was introduced, several works were devoted to
architectural and implementation aspects. For instance,
Krumvieda et al. [55] provided details about analog-to-
digital conversion (ADC), high-sensitivity signal acquisi-
tion, and different tracking loops, and Chakravarthy et al.[56] discussed real-time issues such as the transition from
acquisition to tracking. Textbooks [57] and [58] increased
the awareness of the community about the great benefits
provided by software receivers with respect to the
traditional hardware-oriented approach, providing Matlab
implementations of a complete GPS receiver. In order to
accelerate computations and attain real time in commodity
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1893
general-purpose processors, bitwise operations wereintroduced in [59]. The use of single-input–multiple-data
(SIMD) parallel computing technology for the correlators
and other time-critical operations is due to [60], a solution
that exploited an extension set of assembly instructions for
Intel processors. Both approaches suffered from being bit-
depth dependent, jeopardizing flexibilityVsignal quanti-
fication cannot be easily changed. Other recent approaches
take advantage of today’s pervasive multicore architectureprocessors [61], [62], or of the computational power of
modern graphics processing units (GPUs) [63]. Test
procedures for GNSS software receivers were addressed
in [64], and a general discussion about the architecture is
found in [65]. In [66], Fernandez-Prades et al. advocate the
use of design patterns for the implementation of GNSS
software receivers. Carrier phase measurements and clock
steering are discussed in [67].Today, there are solutions available at academic and
commercial levels, usually not only including program-
ming solutions but also the development of dedicated RF
front-ends. As examples, we can mention the GNSS soft-
ware navigation receiver (GSNRx) developed by the Posi-
tion, Location, and Navigation (PLAN) Group of the
University of Calgary [65], the ipexSR, a multifrequency
(GPS C/A and L2C, EGNOS and GIOVE-A E1-E5a) soft-ware receiver developed by the Institute of Geodesy
and Navigation at the University FAF Munich [68], or
N-Gene, a fully software receiver developed by the
Istituto Superiore Mario Boella (ISMB) and Politecnico
di Torino that is able to process in real time the GPS and
Galileo signals broadcast on the L1/E1 bands, as well as to
demodulate the differential corrections broadcast on the
same frequency by the EGNOS system. This receiver isable to process in real time more than 12 channels, using
a sampling frequency of approximately 17.5 MHz with
8 b/sample [69].
Regarding observable processing and data manage-
ment, the GPS Toolkit (GPSTk) [70], [71] is an open
source project that provides a GNSS computing suite to the
satellite navigation community, consisting of a core lib-
rary, accessory libraries, and some applications. It is alsoworthwhile to mention the NAvigation Package for Earth
Observation Satellites (NAPEOS) software [72], used by
the navigation support office (OPS-GN) at the European
Space Operations Center (ESOC) since January 2008 for
all its International GNSS Service (IGS) activities [73].
C. Signal Processing AlgorithmsThe techniques of digital signal processing play an im-
portant role in the design and implementation of a GNSS
receiver. In fact, GNSS signals are generally very weak.
The receiver thermal noise is dominant over the useful
signal-in-space (SIS), and other signals, such as interfer-
ence and multipath, often impair the weak and fragile line-
of-sight (LOS) SIS received. Since only the LOS SIS
contains the fundamental information about the propaga-
tion delay indicated in (1), the main receiver task is toadopt suitable techniques of statistical signal processing
with the aim of isolating only the contribution carried out
by the LOS SIS from the received signal. The received
signal introduced in (3) can be rewritten as
yðtÞ ¼XNSV
i¼1
rRF;iðtÞ þ NðtÞ þ ðtÞ (45)
where NSV is the number of satellites in view, rRF;iðtÞ is the
received SIS belonging to the ith satellite, NðtÞ is the
thermal noise modeled as a white Gaussian random pro-
cess while ðtÞ includes all the interfering signals at thereceiver antenna. The signals transmitted by GPS, Galileo,
as well as COMPASS, have a code-division multiple-access
(CDMA) format, and GLONASS has also planned to em-
ploy CDMA in the future. For these systems, the received
signal rRF;iðtÞ, after downconversion from RF to interme-
diate frequency (IF) and ADC can be rewritten in terms of
Doppler shift as
rIF½n� ¼ffiffiffiffiffiffiffi2PR
pcbðnTs � �pÞdðnTs � �pÞ cos �ðn; fdÞ (46)
where
�ðn; fdÞ ¼ 2�ðfIF þ fdÞnTs þ ’IF (47)
and Ts is the inverse of the sampling frequency fs, PR is thereceived power, �p is the propagation delay, fIF is the
nominal IF frequency of the receiver front-end, ’IF is a
random phase, and fd is a frequency shift, which includes
both the Doppler shift due to the relative motion between
the satellite and the vehicle and the drift of the local
oscillator. In general, fd varies with time and should be
written as fdðnTsÞ, but in most applications, it can be
considered as a constant. This is because the signal pro-cessing operations performed by the receiver generally
involve data segments where fd does not vary significantly.
The symbol cbðtÞ denotes the signal associated with the
pseudorandom noise (PRN) code of the ith satellite modu-
lated by a subcarrier when present. The navigation mes-
sage is carried by the signal dðtÞ. Notice that a subscript ishould be inserted in the symbols (signals and parameters)
in rIFðnTsÞ to take into account the fact that they refer tothe ith satellite, but the subscript has been omitted here to
simplify the notation.
The ADC introduces a quantization on the IF
discrete-time signal. Most mass-market receivers intro-
duce a 1-b or a 2-b quantization, but with the advent of
SDR technology an 8-b quantization is preferred. The
main reason is that the digital signal can be better seen
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1894 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
in this case as a floating-point discrete-time sequence,for which the results of the digital signal processing and
statistical signal analysis theories can be easily applied.
For this reason, in this section, the IF digital signal will
be written as
yIF½n� ¼ rIF½n� þ wIF½n� (48)
where the term wIF½n� represents a single realization of a
random process WIF½n�, which models the digital version ofthe continuous-time filtered noise at the input of the ADC
block. In most applications, WIF½n� can also include the
contributions of the other satellites (all the satellites in
view except the ith one). This is possible because a CDMA
signal behaves like white noise within the front-end
bandwidth, but the correctness of this model should be
verified case by case. If interference, multipath, and other
disturbing signals are also considered the model in (48)should be modified.
In the GNSS community, the contribution due to noise
is characterized by the carrier-to-noise ratio, generally
expressed in dBHz, defined as
C
N0¼ PR
N0
where PR is the signal power evaluated on the whole signal
bandwidth, while N0 can be considered as the noise power
evaluated over a bandwidth of 1 Hz, or, in other terms, thenoise power density.
The first task of a GNSS receiver is to detect the pre-
sence or absence of a generic satellite. This is done by the
so-called acquisition system, which also provides a coarse
estimation of two SIS parameters: the frequency shift of
the nominal IF frequency, and a delay term which allows
the receiver to create a local code aligned with the incom-
ing code. Since the code is periodic, this means that thisdelay � is a fraction of the code period. The task of the
acquisition system is to provide an estimate � ðAÞ of �
and fðAÞd of fd. These estimated parameters are then fed to
the receiver tracking blocks, which represent the second
stage of the signal processing unit with the aim of
performing a local search for accurate estimates of � and fd.
In this stage, the estimation of the carrier phase may also
be included. Once the signals of the detected satellites are
tracked, the navigation message can be demodulated, thepseudoranges can be measured, and the position, velocity,
and time (PVT) can be evaluated. This is possible because
the information on the propagation delay �p still remains
within the navigation message dðtÞ.The acquisition system is made up of a number of
functional blocks, which conceptually operate indepen-
dently, even if in real systems all or a part of them can be
implemented simultaneously. Two mathematical disci-plines govern the operations performed by an acquisition
system: signal detection theory and estimation theory.
By exploiting the concepts and the methodology of the
estimation theory, it is possible to show that the maximum-
likelihood (ML) estimate of the vector p ¼ ð�; fdÞ, whose
elements are two unknowns of yIF½n�, is obtained by
maximizing the function
pML ¼ arg max�p1
L
XL�1
n¼0
yIF½n��rIF½n������
�����2
(49)
where L is the number of samples in the summation, and�rIF½n� is a test signal, locally generated, of the type
�rIF½n� ¼ cbðnTs � ��Þej2�ðfIFþ�fdÞnTs (50)
where �� and �fd are test variables, defined in a propersupport Dp, which contain all the possible values that can be
assumed by the elements of p ¼ ð�; fdÞ. The summation in
(49) is an inner product, known as cross-ambiguity func-
tion (CAF). In fact, by writing the summation in (49) as
Ry;rð�pÞ ¼1
L
XL�1
n¼0
yIF½n�cbðnTs � ��Þej2�ðfIFþ�fdÞnTs (51)
the structure of an ambiguity function can be clearly
recognized. The result of (49) holds only if the energy of
the test signal �r½n� does not depend on �p ¼ ð��; �fdÞ. A proof
of this result can be found in [74], where it is also shownthat (49) is not sensitive to the phase term ’IF, from which
the name noncoherent acquisition scheme is taken, used to
identify an acquisition engine based on (49). Finally, no-
tice that the presence of data bits can impair the estimation
process if not properly taken into account.
The CAF is also used to decide if a specific satellite is in
view or not. The decision variable
Smax ¼ max�p Ry;rð�pÞ�� ��2 (52)
is compared against a threshold t to test two possiblehypotheses
H1 : The satellite is in view
H0 : The satellite is not in view.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1895
If Smax > t, the satellite is declared present, otherwise itis considered absent. The performance of the detector is
evaluated in terms of detection probability Pd ¼ PðSmax >tjH1Þ and false alarm probability Pfa ¼ PðSmax > tjH0Þ.Notice that if different strategies are used to perform the
detection, the two probabilities Pd and Pfa should be de-
fined differently. This topic is dealt with in [75]. Other
strategies are also possible. For example, in [76], a method
based on Bayesian sequential detection is described.The CAF evaluation is generally very time consuming,
as a huge matrix of points has to be evaluated before
applying the estimation and detection processes. One way
to reduce the complexity of these operations is to adopt the
so-called serial search. With this approach the sequence of
the two operations (CAF evaluation and detection) is
interrupted the first time a value crosses the threshold t.
This is a suboptimal method, as it does not guarantee toobtain the ML solution, based on the CAF maximum.
Nonetheless, it is evident that this approach is much faster.
Other methods exist based on block processing techniques,
which allow the evaluation of an entire row or column of
the CAF matrix in a single shot. These methods exploit
some interesting properties of the CAF. In fact, if the
signals involved in (51) are grouped appropriately, the
presence of either a correlation or a Fourier transforminside the CAF structure becomes evident. As a conse-
quence, it is clear that the fast Fourier transform (FFT),
the most famous signal processing algorithm, can play a
key role in speeding up the CAF evaluation. In fact, both
the discrete-time Fourier transform and the correlation
can be implemented with very efficient and fast algorithms
based on FFT.
As previously anticipated, it is the task of the trackingblocks to fine estimate the code delay � and frequency shift
fd. This objective is performed by synchronizing the code
and the carrier of the received signal with local generated
signals. Although this task could potentially be pursuedworking with the CAF on a bi-dimensional space, in con-
ventional receivers this signal alignment is obtained by
means of two mono-dimensional algorithms:
• the code delay synchronization is performed by
means of delay lock loops (DLLs);
• the frequency shift synchronization is obtained by
means of frequency lock loops (FLLs) or the whole
carrier phase is recovered by means of phase lockloops (PLLs).
The operations of the FLL, PLL, and DLL are regulated by
the same functioning and they differ only on the signal
which is to be synchronized. The DLL needs to operate on
a signal where the carrier is wiped off in order to process a
clean PRN sequence, while FLL and PLL require a signal
where the code is wiped off.
Wipe-off and synchronization can be done by means of a
concatenated scheme, as shown in Fig. 2, where two wipingsystems are coupled together and linked to the estimators
of code delay and frequency shift. At each iteration the
estimators provide fresh estimates of frequency shift and
code delay to the wiping systems, which progressively
improve the quality of the signals at their outputs.
The fine synchronization of the code can be obtained
by using the fundamental properties of the PRN code
correlation function. Correlation is an even function whichassumes its maximum only when the signal and its replica
are perfectly aligned, and therefore, the synchronization
issue becomes a maximization problem. Moreover, recal-
ling that the derivative of a function is null at a minimum
and maximum value, the problem can be transformed
again to become a problem of null searching. Similar con-
siderations are also valid for the fine estimation of the
frequency shift.This is why DLL, FLL, and PLL are generally referred to
as null-seeker systems. In fact, their main purpose is to
Fig. 2. Concatenation of wipe-off and tracking systems.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1896 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
find a zero, and this null search is pursued in an iterativeway. At each iteration, the correlation function is eval-
uated, and transformed, if necessary, in another function
without altering the position of the maximum. The next
step is the evaluation of a derivative (appropriately ap-
proximated), and the search of its zero point. The loop is
closed by applying a low-pass filter in order to reduce
the contribution of the noise from the measurement as
much as possible without limiting the reactivity of theloop.
Although null-seeker schemes are well established, the
search for the optimal discrimination function in GNSS
applications is not over yet. In fact, the performances of
the tracking loops greatly depend on the shape of the cor-
relation function, and, as a consequence, on the discrim-
ination function being adopted. Two key performance
parameters for the tracking loops are:• tracking jitter that characterizes the quality with
which the signal is synchronized over the time;
• multipath rejection or the error committed by the
loop when the signal is received over multiple
directions of arrival.
Several proposals can be found in the literature for
the discriminator function, which aim to reduce jitter and
to increase the capability to reject multipaths. An impor-tant aspect to be taken into account in the definition of
these functions is the innovation introduced by the pre-
sence of BOC subcarriers in the new GNSS signals. These
subcarriers lead to narrower main peaks and introduce
side lobes, as seen in Fig. 3(a), which shows three auto-
correlation functions obtained with ideal waveforms: a
rectangular chip for a GPS C/A signal, a BOC(1,1) and a
BOC(10,5). These characteristics are very robust and canbe also observed in the real signals, as shown in Fig. 3(b),
where the cross correlation between a local ideal code
and the incoming signal is drawn in the presence of noise
(C=N0 ¼ 42 dBHz) and front-end filtering. A complete
description of the possible discriminator functions
associated to the new autocorrelation functions is beyond
the scope of this paper; some examples can be found
in [77].
VI. HYBRIDIZATION WITHINERTIAL SENSORS
Inertial navigation has a long history, beginning between
the World Wars I and II and evolved principally for
military purposes, namely for the guidance of rockets,
spacecrafts, guided missiles, and then civil aircrafts.Inertial navigation is based on two families of inertial
sensors: accelerometers and gyroscopes. Three orthogonal
accelerometers rigidly mounted on a body provide mea-
surements of acceleration in a 3-D frame. Formally, cas-
caded integrations over time of such measurements yield
a measure of the instantaneous velocity and position of
the body itself. However, as long as the accelerometers
are rigidly attached to the body in movement (strapdownsystem), their reference frame is the body frame (i.e., a
frame integral with the body), which is usually meaning-
less in order to provide an indication of the body move-
ment with respect to an external reference frame (e.g.,the Earth). Therefore, the angular rotation of the body
with respect to the external reference frame (attitude),
must be computed from the instantaneous angular
orientation of the body. This is obtained by integrating
the instantaneous turn rate the body is subject to, mea-
sured by three gyroscopes rigidly mounted on the body
along the orthogonal accelerometers axes. Thereby, the
attitude information is used to resolve the accelerometermeasurements into the external reference frame [78],
[79]. A set of inertial sensors (accelerometers and gyros-
copes, plus sometimes a magnetometer and/or an
altimeter) forms an inertial measurement unit (IMU).
Fig. 3. Comparison of correlation functions. (a) Ideal
autocorrelation functions of GPS C/A code, BOC(1,1) and BOC(10,5).
(b) Cross-correlation functions for GPS C/A and Galileo BOC(1,1) signals,
in presence of noise (C=N0 ¼ 42 dBHz) and front-end filtering.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1897
The IMU coupled with a computational unit to resolve thesystem mechanization is an inertial navigation system
(INS). A comprehensive treatment of strapdown INSs can
be found in [52], [79], and [80].
The goal of integrating INSs with GNSS is to take
advantage of the complementary characteristics of the two
systems. The rate of the solution output by an INS is
usually higher than that offered by a GNSS receiver. INSs
experience relatively low-noise solutions, but they tend todrift over time. The reason is the fact that body position
and velocity are obtained from double and single integra-
tion, respectively, of acceleration measurements; simi-
larly, angular rotations are obtained from turn rate
integration. Therefore, the errors on the INS-estimated
trajectory are potentially unbounded. For these reasons,
INS performance strongly depend on the quality of the
sensors, in terms of bias on accelerometers and gyroscopesand noise. On the other hand, a GNSS receiver produces a
solution affected by an error always bounded by the char-
acteristics of the receiver itself, but with an error variance
prone to the propagation nuisances, as discussed in the
previous sections; furthermore, in particular conditions,
the satellite signals could be unavailable for the receiver
(e.g., in urban canyons, tunnels, under dense foliage,
under water, etc.), thus causing an outage of the solution.Additionally, opposite to INSs, GNSSs are sensitive to
jamming. The final goal of integrating the two systems is,
therefore, to improve their performance in those condi-
tions when one system alone would fail to work, or would
work with poor performance.
The literature identifies three conceptual approaches
for integrating INS-based positioning and GNSS-based
positioning (system hybridization): 1) loose integration;2) tight integration; and 3) ultratight integration. These
solutions differ for the degree of integration of the two
systems, i.e., for the nature of the information extracted
from the two systems and used in the hybridization pro-
cess, as well as for the architecture of their interactions.
System hybridization is basically a mathematical matter,
usually stated through a state-space model and resolved
with a Kalman filter (KF), in one of its several variants[81]–[83]. Although other mathematical tools for the sys-
tem hybridization can be envisaged (e.g., open-loop batch/
sequential processing [84], particle filtering [85], [86],
neural networks [87]), the KF is historically and concep-
tually the principal approach, thanks to its ability to blend
different sources of noisy measurements in a single state-
space description of the system evolution. The most
common approach cited in the literature relies on anincremental state-space model associated to an extended KF
(EKF) algorithm. The states are defined as the difference
(increment) between the true state (what we need to
estimate) and a nominal evolution of the states (nominaltrajectory), determined in some independent way.
Without entering into derivation details, which may be
found in [88] and [89], the discrete-time state-space model
can be written as
�x½nþ 1� ¼ %½n��x½n� þw½n� (53)
where �½nþ 1� is the incremental state vector at time
nþ 1, %½n� is the state transition matrix, which relates the
states at the time n to the states at the time nþ 1, and w½n�is the time-uncorrelated model noise, with known
statistics. The state vector �½n� typically contains, at least
[88], [89]: the 3-D incremental position (i.e., the correc-tions vector to be applied to the nominal body position at
the instant n); the 3-D incremental velocity; the incre-
mental attitude; the vector of the accelerometers’ biases;
the vector of the gyroscopes’ biases; and, possibly, the re-
ceiver clock’s bias and drift. Note that this basic structure
sums up to 17 1-D states.
The state transition matrix %½n� rules the joint evolu-
tion of the states, so it is written in terms of directioncosine matrices (from the body reference frame to the
inertial one), angular rotation matrices, effects of the gra-
vitational and centripetal accelerations, and the mathe-
matical models of the inertial sensor biases and drifts [89].
The measurement equation of the EKF has the usual
structure
�z½n� ¼ H½n��x½n� þ N½n� (54)
where �z½n� is the incremental observation vector andH½n� is the observation matrix. The term N½n� is an additive
noise component with know statistical properties. �z½n� isdefined as the difference between a vector of measure-
ments taken from the real system and a vector of nominalmeasurements extracted from the nominal state trajectory.
In GNSS-INS hybridization, the nominal state trajectory is
due to the INS [89], [90]. However, the kind of system
observables contained in the vector �z½n� vary with theintegration architecture, as well asVconsequentlyVthe
structure of the observation matrix H½n� and the statistics
of the measurement noise N½n�.
A. Loose IntegrationIn a loosely integrated architecture, the GNSS receiver
and the INS act as independent navigation systems. Their
navigation solutions (body position and velocity) are then
blended in order to obtain a better estimate of the body
trajectory. The conceptual block diagram of a looselycoupled architecture is shown in Fig. 4. The system ob-
servables are the body positions and velocities as com-
puted, separately, by the INS and the GNSS receiver. Their
difference is taken so as to define the observation vector of
the EKF; therefore, six 1-D variables define the vector
�z½n� (the so-called nominal measurements are repre-
sented by the INS-predicted position and velocity). Note
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1898 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
that this vector can be computed only when both trajectory
solutions are available; but typically the solution rate of the
INS is higher than that of the GNSS receiver. Furthermore,in case of GNSS outage due to environmental conditions
(e.g., indoor navigation) the GNSS solution could be un-
available for relatively long periods of time. In such cases
the navigation could be demanded to the INS only.
B. Tight IntegrationA tightly integrated system uses the pseudorange and
pseudorange rate information extracted from the GNSS
receiver to compute the corrections to be applied to thetrajectory estimated by the INS computer and to estimate,
if necessary, the biases that affect the accelerometers and
the gyroscopes. The GNSS information is used as a refine-
ment of the INS information, so as to counteract the in-
trinsic derivation of the INS solution, correcting the INS
trajectory. The architecture of a tight integration is repre-
sented in Fig. 5: the INS-predicted trajectory is employed
to predict the pseudoranges of all visible satellites, thenproviding the nominal measurements. Thus, the EKF uses
the differences in pseudoranges and pseudorange rates for
each visible satellite as its observations; in this way, theinformation available to bound the INS errors is in some
sense proportional to the number of satellites in view.
C. Ultratight IntegrationUltratight hybridization architectures are a more re-
cent technology, on which different approaches may
coexist and confront (e.g., [91]–[98]). In ultratight hybri-
dization, GNSS and INS interact at the Bloop-filter level,[computing the runtime corrections of the Doppler fre-
quency and code phase by means of the hybridization
filter, thus refining (or even replacing) the PLL’s and DLL’s
estimations. This approach promises to improve the re-
ceiver performance in tracking the signal dynamics at the
antenna and to be robust is case of fading, strongly atte-
nuated signals, high dynamics, and even GNSS outages,
thanks to the additional information provided by the INS[91], [97], [99], [100].
The underlying idea is to predict the signal observed at
a certain stage of the receiving chain (typically, after the
correlators stage) using an INS-based prediction of the
receiver motion to estimate the Doppler frequency.
Coherent ultratight architectures measure the received sig-
nal on the in-phase and quadrature output of each chan-
nel’s prompt correlator, plus, sometimes, the early and lateones [91], [92], [97]. Noncoherent architectures, on the
other hand, observe either the output of the code and
carrier/phase discriminators or the received pseudoranges
[92], offering a way to realize ultratight hybridization
using a microelectromechanical systems (MEMS)-based
IMU [101]. The blending filter then predicts the amount of
Doppler frequency shift the signal currently experiences,
thereby aiding the PLL and the DLL in tracking the signaleven in high noise/high fading/high dynamics conditions.
Additional states may be the signal amplitude, the code
phase tracking error, the phase tracking error, the fre-
quency tracking error, and the frequency rate error of the
numerically controlled oscillator [94]. An innovative ar-
chitecture where the usual EKF is replaced by a square root
cubature Kalman filter has been recently proposed in [98].
VII. MASS-MARKET PERSPECTIVE
The growing need for ubiquitous positioning and naviga-
tion to serve the Banywhere, anytime[ needs of the con-
sumer, the LBS and the ITS markets will strongly influence
future mass-market receiver architectures. BAnywhere,
anytime[ navigation requirements will favor architectures
that can deal with a large variety of position-related infor-mation. Rather than relying on just one signal type for
navigation, future receivers may need to scan through the
universe of potentially available position information
sources (e.g., a multiplicity of GNSS systems, but also
wireless communication systems and IMUs), and dynam-
ically choose those that are available at its current location
and time.
Fig. 4. Block diagram for a GNSS-INS loosely integrated architecture.
Fig. 5. Block diagram for a GNSS-INS tightly integrated architecture.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1899
Satellite positioning is expected to pervade most ofportable devices, enabling an ever growing number of
location-aware applications for the mass market such as
social networking, local search, geotagging, street map and
direction finding, augmented reality, fleet position data
logging, location-based marketing and targeted advertising,
leisure support (tourism, hiking, biking, etc.), messaging, or
microblogging. This shift from dedicated devices and
applications to Blocation as an embedded feature[ will resultin an integrated environment where location experience
becomes ubiquitous, seamless, and transparent to the user.
GNSS receivers are becoming a standard feature not only on
medium-high end mobile phones, but also on other portable
devices such as digital cameras and portable gaming
consoles. Mobile positioning for E911 and E112 emergency
services is becoming more pervasive, and a plethora of user
applications based on mobile location is rapidly emerging.An example could be Apple’s iPhone (and its associated
operating system iOS): each version adds new navigation
capabilities with respect to the previous one. WiFi position-
ing, cellular tower triangulation, GPS, and compass were
added in this order in successive iPhone versions. Bestselling
applications are related to navigation, and they are not the
cheapest [102]. This enhancement of LBS capabilities has
resulted in a rapid settlement in the smartphone market.This poses challenges to the whole value chain: re-
searchers, chipset manufacturers, device vendors, map
data providers, application stores, developers, and mobile
operators. The impact of interoperability, GNSS-based
management of the wireless network infrastructure from
the perspective of network capacity and quality of service,
location in social networks, and privacy issues are aspects
to be further analyzed.The hugely wide variety of application fields for GNSS
mentioned above pushes mass-market GNSS receivers to
face an increasing demand for higher positioning avail-
ability and accuracy particularly in urban and indoor
environments, where GNSS signals are greatly attenuated
or even blocked. The answer to this challenge is coming
from two technologies that faced an explosive develop-
ment in the last ten years: high-sensitivity1 (HS) GNSS andassisted GNSS. Their adoption has substantially improved
the performance of GNSS receivers operating in critical
propagation environments and has also become a common
feature in commercial low-end receivers. Assisted GNSS,
for example, is offered today on most commercial general-
purpose GPS chips. Other augmentation techniques, less
standardized and still farther from being widely employed
in mass-market devices, may exploit, for example, theinformation extracted from inertial sensors, odometers,
barometric altimeters, vision, WiFi localization, and ultra-
wideband ranging [99], [104]. Although commercial-grade
indoor navigation can be currently handled with a tech-nology mix (usually blending GNSS with cellular, RFID,
inertial sensors, and wireless communication technolo-
gies), there is still a lack of well-established procedures
and standards for positioning in environments with no
LOS within the receiver and the satellites, and it consti-
tutes an active field of research and development.
The nominal minimum GPS signal strength for a user
on the earth surface and with unobstructed visibility to thesatellite is defined as �130 dBm for the C/A code [9]. This
level is increased to �127 dBm for Galileo E1 [13]. LOS
obstructions may attenuate the received power by about
5 dB in cars, up to 20 dB in buildings, and more that
25 dB in subterranean garages [105], severely degrading
the signal processing performance of the receiver.
HS receivers are recognized to have Brevolutionized the
GPS receiver market[ [106] by extending the GPS serviceavailability and allowing integration of GPS chips into
personal handsets, such as mobile phones, equipped with
cheap and low performing antennas. For years, most providers
have sold HS receivers that acquire signal below �150 dBm,
although the high-sensitivity regime should be assumed to
refer to signal levels of �155 dBm and below [103].
Weak signals, attenuated by windows, walls, roofs, fur-
niture, foliage, and so on, usually undergo further impair-ments, typical in urban canyons and indoor conditions:
multipath, cross-correlation false locks, and the squaring
losses [100], [107]. The fundamental approach to deal with
such problems is increasing the coherent signal integration
time, as the signal-to-noise ratio of weak signals is expected
to improve proportionately to the coherent integration
time. In particular, a coherent integration time of several
seconds would mitigate the main indoor impairments.HS receivers usually extend the coherent integration time
up to the length of one GPS navigation data bit, corresponding
to a time of 20 ms. Nonetheless, a significant gain could be
expected by applying a much longer coherent integration.
However, various aspects make the use of very long coherent
integration difficult. Apart from the data bit transitions,
which occur every 20 ms in GPS and every 4 ms in the
data channel of Galileo, and which can be predicted usingassistance data from a reference station with open sky
condition, the internal clock of low-cost devices is typically
not stable enough to provide sufficient jitter accuracy for
generating the local replica of the code and carrier.
Therefore, a stable oscillator, e.g., an OCXO, is necessary
to reduce oscillator jitter to acceptable values [100].
Today, typical values of achievable signal acquisition
sensitivity for GPS C/A can be set to around �157 dBm or17 dBHz, assuming a �174-dBm/Hz noise power density.
These values have been found by Pany et al. [106] using an
experimental setup claimed to be Brepresentative of the
state of the art in the year 2010,[ based of the IFEN
INTrack ASIC and the SX Navigation Software Receiver.
On the other hand, nominal values of acquisition and
tracking sensitivity for three commercial HS receivers are
1Sensitivity is defined as the lowest signal power, in dBm, detectableby a certain receiver. Sensitivity should be measured in the context of aspecific receiver operation mode, e.g., acquisition sensitivity, trackingsensitivity, and re-acquisition sensitivity [103].
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1900 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
reported in Table 1. A very challenging aspect in increasing
sensitivity via long coherent integrations, however, is the
reproduction of the user motion with accuracy high enoughto compensate the nonlinear dynamics on the pseudorange
and the Doppler frequency. Various approaches proposed
in the literature, which are likely to be more suitable to
high-end/professional receivers, are based on vector
tracking [99], [108]. Vector-based tracking architectures
aim at improving the signal tracking performance of GNSS
receivers by using the position, velocity, and clock esti-
mates from the navigation filter to predict the signal char-acteristics from each satellite (code phase and Doppler
frequency). In this way, the receiver can better predict the
incoming signal characteristics, thereby allowing the use of
a narrower tracking loop bandwidth. The narrower band-
width then reduces the amount of noise entering the track-
ing loops and allows the vector-based receiver to track
weaker signals in challenging signal environments. Examples
of such an approach are reported in [99] and [100]. Anarchitecture that couples both long (noncoherent) integra-
tions and high-sensitivity tracking loops is proposed in [109].
Increasing receiver sensitivity and service availability,
as well as coverage, for mass-market applications is at-
tracting so much interest that even more Bfuturistic[ ap-
proaches are deserving particular attention nowadays. One
of such innovative approaches, particularly promising in
terms of performance, is the so-called peer-to-peer posi-tioning (P2P). P2P assumes the existence of a wireless
point-to-point mobile network, where nodes (users) can
exchange position-related data, in order to perform coope-
rative positioning and/or timing. A typical situation where
P2P positioning could be foreseen is a Vehicular Ad-hoc
NETwork (VANET) [110], [111]. The wireless technology
enabling the implementation of such a network is expected
to be the wireless access in vehicular environment(WAVE), a communication standard designed for rapidly
changing propagation environments and also where very
short-duration short-range data exchanges are required.
Cooperative positioning methods can be classified in
two families: methods based on the exchange of GNSS-
data only among peers and hybrid methods, based on bothGNSS data and other positioning measurements (e.g., ter-
restrial ranging, inertial sensors, odometers, etc.). In the
context of exchanging GNSS-data only, several procedures
can be envisaged to implement the concept of P2P posi-
tioning. A first class of P2P procedures refers to techniques
related to the physical layer of the signal (in particular,
aided acquisition approaches, such as [112] and [113]).
Other techniques work at the pseudorange layer (PVTcomputation) and exploit the position computed by the
aiding peers to evaluate the position of the aided peer
[114]. Finally, the peers in open sky condition can behave
as pseudosatellites, which can be used by the target peer to
fix its position. Interestingly, these methods feature bene-
fits with respect to the usual assistance approaches (e.g.,
A-GNSS, HS standalone GNSS receivers), thanks to the
vicinity of the peers involved in the cooperation and, at thesame time, to the diversity provided by the physical
separation of the involved receivers [115].
Today, location is also a key aspect in fields such as
aviation [116], ITS (addressing the problems of good trace-
ability and tracking, road safety, pollution, and conges-
tion), and high-precision applications such as precision
agriculture, time transfer, or surveying. Dramatic improve-
ments of performance are envisaged in the following years,continuing to spin the circle of science, technology, and
business around navigation.
VIII . CONCLUSION
The modernization of existing GNSSs and the advent of new
ones dramatically change the landscape of civil positioning
systems, enabling new applications and business models but
posing challenges in the design of receivers able to fully
exploit the available signals. This paper described satellite-
based navigation systems and signals, augmentation systems,and receiver technology that will play an important role in
future positioning systems, from a mass-market perspective.
We also identified further research topics in the design of
multiconstellation, multifrequency, high-sensitivity recei-
vers, and the integration and interoperability with other
technologies, such as inertial navigation systems and
wireless, ground-based communication technologies. h
Acknowledgment
The authors would like to thank A. Ordenes and
M. McIntosh for their manuscript review.
RE FERENCES
[1] D. Sobel, Longitude. The True Story of a LoneGenius Who Solved the Greatest ScientificProblem of His Time. London, U.K.: Fouth Estate, 1995.
[2] J. Stone, BMethod of determining thedirection of space telegraph signals,[ Boston,MA, U.S. Patent 716,134, Dec.16, 1902.
[3] M. Hernandez-Pajares, J. M. J. Zornoza,J. S. Subirana, R. Farnworth, and S. Soley,BEGNOS test bed ionospheric correctionsunder the October and November 2003storms,[ IEEE Trans. Geosci. RemoteSens., vol. 43, no. 10, pp. 2283–2293,Oct. 2005.
[4] S. Bancroft, BAn algebraic solution ofthe GPS equations,[ IEEE Trans. Aerosp.Electron. Syst., vol. 21, no. AES-1, pp. 56–59,Jan. 1985.
[5] G. Strang and K. Borre, Linear Algebra,Geodesy, and GPS. Wellesley, MA:Wellesley-Cambridge Press, 1997.
Table 1 HS Performance of Three Commercial Receivers, as Declared by
the Manufacturers’ Datasheets
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1901
[6] P. J. G. Teunissen, BThe least-squaresambiguity decorrelation adjustment:A method for fast GPS integer ambiguityestimation,[ J. Geodesy, vol. 70, no. 1–2,pp. 65–82, 1995.
[7] W. Werner and J. Winkel, BTCAR andMCAR options with Galileo and GPS,[ inProc. 16th Int. Tech. Meeting Sat. Div.Inst. Navig., Portland, OR, Sep. 2003,pp. 790–800.
[8] E. D. Kaplan and C. J. Hegarty, Eds.,Understanding GPS. Principles andApplications, 2nd ed. Norwood, MA:Artech House, 2006.
[9] Science Applications InternationalCorporation, Interface SpecificationIS-GPS-200 Revision E. Navstar GPSSpace Segment/Navigation User Interfaces,El Segundo, CA, Jun. 2010.
[10] Science Applications InternationalCorporation, Interface SpecificationIS-GPS-800 Revision A. Navstar GPS SpaceSegment/User Segment L1C Interfaces,El Segundo, CA, Jun. 2010.
[11] Science Applications InternationalCorporation, Interface SpecificationIS-GPS-705 Revision A. Navstar GPS SpaceSegment/User Segment L5 Interfaces,El Segundo, CA, June 2010.
[12] Russian Institute of Space DeviceEngineering, Global Navigation SatelliteSystem GLONASS. Interface Control Document.Navigational Radiosignal in Bands L1, L2,Moscow, Russia, 2008, ver. 5.1.
[13] European Union, European GNSS (Galileo)Open Service. Signal In Space Interface ControlDocument. Ref: OS SIS ICD, Issue 1.1,Sep. 2010.
[14] N. Gerein, BA hardware architecture forprocessing alternate binary offset carrier(AltBOC) signals,[ Paris, France, EuropeanPatent EP 1 644 753 B1, Dec. 2007, EuropeanSpace Agency.
[15] T. Grelier, J. Dantepal, A. DeLatour,A. Ghion, and L. Ries, BInitial observationsand analysis of Compass MEO satellitesignals,[ Inside GNSS, vol. 2, no. 4, pp. 39–43,May/Jun. 2007.
[16] G. X. Gao, A. Chen, S. Lo, D. D. Lorenzo, andP. Enge, BGNSS over China: The CompassMEO satellite codes,[ Inside GNSS, vol. 2,no. 5, pp. 36–43, Jul./Aug. 2007.
[17] G. X. Gao, A. Chen, S. Lo, D. D. Lorenzo,T. Walter, and P. Enge, BCompass-M1broadcast codes in E2, E5b, and E6 frequencybands,[ IEEE J. Sel. Topics Signal Process.,vol. 3, no. 4, pp. 599–612, Aug. 2009.
[18] L. L. Weill, BMultipath mitigation. How goodcan it get with new signals?[ GPS World,pp. 106–113, Jun. 2003.
[19] D. Gabor, BTheory of communication,[J. Inst. Electr. Eng., vol. 93, no. 3,pp. 429–457, 1946.
[20] T. Richert and N. El-Sheimy, BOptimal linearcombinations of triple frequency carrierphase data from future global navigationsatellite systems,[ GPS Solutions, vol. 11,no. 1, pp. 11–19, Jan. 2007.
[21] P. Closas and C. Fernandez-Prades,BBayesian nonlinear filters for direct positionestimation,[ in Proc. IEEE Aerosp. Conf.,Big Sky, MT, Mar. 2010, DOI: 10.1109/AERO.2010.5446676.
[22] P. Closas, C. Fernandez-Prades, andJ. A. Fernandez-Rubio, BCramer-Rao boundanalysis of positioning approaches in GNSSreceivers,[ IEEE Trans. Signal Process.,vol. 57, no. 10, pp. 3775–3786,Oct. 2009.
[23] Japan Aerospace Exploration Agency,Quasi-Zenith Satellite System NavigationService. Interface Specification for QZSS(IS-QZSS), Draft V1.2, Mar. 2010.
[24] K. N. S. Rao, BPresentation of the IndianRegional Navigation Satellite System(IRNSS),[ in Proc. 2nd Meeting Int.Committee Global Navig. Satellite Syst.,Bangalore, India, Sep. 2007. [Online].Available: http://www.oosa.unvienna.org/oosa/SAP/gnss/icg/icg02/presentations.html.
[25] P. Majithiya, K. Khatri, and J. K. Hota,BIndian regional navigation satellite system.Correction parameters for timing groupdelays,[ Inside GNSS, vol. 6, no. 1, pp. 40–46,Jan./Feb. 2011.
[26] T. Walter, P. Enge, J. Blanch, and B. Pervan,BWorldwide vertical guidance of aircraftbased on modernized GPS and new integrityaugmentations,[ Proc. IEEE, vol. 96, no. 12,pp. 1918–1935, Dec. 2008.
[27] RTCA, Minimum Operational PerformanceStandards for Global Positioning System/WideArea Augmentation System AirborneEquipment, DO-229D, Washington, DC,Dec. 13, 2006.
[28] P. Enge, T. Walter, S. Pullen, C. Kee,Y. C. Caho, and Y. J. Tsai, BWide areaaugmentation of the Global PositioningSystem,[ Proc. IEEE, vol. 84, no. 8,pp. 1063–1088, Aug. 1996.
[29] S. S. Jan, T. Walter, and P. Enge, BReversionfrom L1-L5 dual to L5 single frequencyWAAS in the presence of RF interference,[IEEE Trans. Aerosp. Electron. Syst., vol. 46,no. 3, pp. 1110–1126, Jul. 2010.
[30] J. Ventura-Traveset and D. Flament, Eds.,EGNOS. The European GeostationaryNavigation Overlay System–A cornerstoneto Galileo. Noordwijk, The Netherlands:European Space Agency, Dec. 2006.
[31] EGNOS Service Definition Document OpenService. Ref: EGN-SDD OS, Revision 1.0,European Commission. Directorate-Generalfor Energy and Transport, Oct. 1, 2009.
[32] EGNOS Safety of Life Service DefinitionDocument. Ref: EGN-SDD SoL, Revision 1.0,European Commission. Directorate-Generalfor Enterprise and Industry, Mar. 3, 2011.
[33] A. D. Sarma, D. V. Ratnam, and D. K. Reddy,BModelling of low-latitude ionosphere usingmodified planar fit method for GAGAN,[IET Radar Sonar Navig., vol. 3, no. 6,pp. 609–619, Dec. 2009.
[34] K. Dixon, BStarFire: A global SBAS forsub-decimeter precise point positioning,[ inProc. 19th Int. Tech. Meeting Sat. Div. Inst.Navig., Fort Worth, TX, Sep. 2006,pp. 2286–2296.
[35] L. E. Ott, BThe OMNISTAR virtual basestation system,[ in Proc. Position LocationNavig. Symp., Atlanta, GA, April 2006,pp. 590–595.
[36] M. Sahmoudi, J.-L. Issler, F. Perozans,Y. Tawk, A. Jovanovic, C. Botteron,P.-A. Farine, R. J. Landry, V. Dehant,A. Caporali, S. Reboul, and P. Willis,BU-SBAS: A Universal multi-SBAS standardto ensure compatibility, interoperabilityand interchangeability,[ in Proc. 5thESA Workshop Satellite Navig. Technol.,Noordwijk, The Netherlands, Dec. 2010,pp. 1–18, DOI: 10.1109/NAVITEC.2010.5708076.
[37] T. Murphy and T. Imrich, BImplementationand operational use of ground-basedaugmentation systems (GBASs)VAcomponent of the future air trafficmanagement system,[ Proc. IEEE, vol. 96,no. 12, pp. 1936–1957, Dec. 2008.
[38] M. Hernandez-Pajares, J. M. Juan, J. Sanz,A. Aragon-angel, P. R. Bosch, J. Samson,M. Tossaint, M. Albertazzi, D. Odijk,P. J. G. Teunissen, P. de Bakker, S. Verhagen,and H. van der Marel, BWide-Area RTK.High precision positioning on a continentalscale,[ Inside GNSS, vol. 5, no. 2, pp. 35–46,Mar./Apr. 2010.
[39] F. van Diggelen, A-GPS: Assisted GPS, GNSS,and SBAS. Norwood, MA: Artech House,2009.
[40] L. Wirola, BPositioning protocol for next-gencell phones,[ GPS World, vol. 22, no. 3,pp. 8–13, Mar. 2011.
[41] S. M. Radicella, BThe NeQuick modelgenesis, uses and evolution,[ Ann.Geophys., vol. 52, no. 3/4, pp. 417–422,Jun./Aug. 2009.
[42] T. Kadoyama, N. Suzuki, N. Sasho, H. Iizuka,I. Nagase, H. Usukubo, and M. Katakura,BA complete single-chip GPS receiver with1.6-V 24-mW radio in 0.18-�m CMOS,[IEEE J. Solid-State Circuits, vol. 39, no. 4,pp. 562–568, Apr. 2004.
[43] P. G. Mattos, BA single-chip GPS receiver,[GPS World, vol. 16, no. 10, pp. 24–30,Oct. 2005.
[44] NaviLink 6.0 single-chip A-GPS/Bluetooth/FMRx/Tx solution: NL5500, Texas Instruments,Dallas, TX, Product Bulletin, Feb. 2009.
[45] GPS IC Market Vendor Matrix, ABI Research,Oyster Bay, NY, March 2009.
[46] G. Artaud, L. Ries, J. Dantepal, J. L. Issler,T. Grelier, A. Delatour, O. Julien, andC. Macabiau, BCBOC performances usingsoftware receiver,[ in Proc. 2nd WorkshopGNSS Signals Signal Process., Noordwick,The Netherlands, Apr. 2007, pp. 1–12.
[47] P. Paimblanc, W. Chauvet, D. Bonacci,T. Sadiki, and F. Castanie, BImprovedpositioning using GSM and GNSS tighthybridization,[ in Proc. Eur. Navig. Conf.Global Navig. Satellite Syst., Toulouse, France,Apr. 2008, pp. 1–7.
[48] K. Pahlavan, F. Akgul, Y. Ye, T. Morgan,F. Alizadeh-Shabdiz, M. Heidari, andC. Steger, BTaking positioning indoors.Wi-Fi localization and GNSS,[ Inside GNSS,vol. 5, no. 3, pp. 40–47, May 2010.
[49] R. Shibasaki, K. Chino, and D. Manandhar,BSeamless tracking: GNSS plus RFID,[GPS World, vol. 20, no. 8, pp. 30–34,Aug. 2009.
[50] S. Gezici, Z. Tian, G. B. Giannakis,H. Kobayashi, A. F. Molisch,H. V. Poor, and Z. Sahinoglu, BLocalizationvia ultra-wideband radios: A look atpositioning aspects for future sensornetworks,[ IEEE Signal Process. Mag.,vol. 22, no. 4, pp. 70–84, Jul. 2005.
[51] G. Mao and B. Fidan, Eds., LocalizationAlgorithms and Strategies for Wireless SensorNetworks. Hershey, PA: InformationScience Reference (IGI Global), 2009.
[52] M. S. Grewal, L. R. Weill, and A. P. Andrews,Global Positioning Systems, Inertial Navigation,and Integration, 2nd ed. Hoboken, NJ:Wiley, 2007.
[53] H. Hurskainen, E.-S. Lohan, J. Nurmi,S. Sand, C. Mensing, and M. Detratti,BOptimal dual frequency combination forGalileo mass market receiver baseband,[ inProc. IEEE Workshop Signal Processing Syst.,Tampere, Finland, Oct. 2009, pp. 261–266.
[54] D. Akos, BA software radio approach toglobal navigation satellite system receiverdesign,[ Ph.D. dissertation, College Eng.Technol., Ohio Univ., Athens, OH,Aug. 1997.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1902 Proceedings of the IEEE | Vol. 99, No. 11, November 2011
[55] K. Krumvieda, P. Madhani, C. Cloman,E. Olson, J. Thomas, P. Axelrad, andW. Kober, BA complete IF software GPSreceiver: A tutorial about the details,[ inProc. 14th Int. Tech. Meeting Sat. Div. Inst.Navig., Salt Lake City, UT, Sep. 2001,pp. 789–829.
[56] V. Chakravarthy, J. Tsui, D. Lin, andJ. Schamus, BSoftware GPS receiver,[GPS Solutions, vol. 5, no. 2, pp. 63–70,Oct. 2001.
[57] J. B.-Y. Tsui, Fundamentals of GlobalPositioning System Receivers. A SoftwareApproach, 2nd ed. New York: Wiley, 2005.
[58] K. Borre, D. M. Akos, N. Bertelsen,P. Rinder, and S. H. Jensen, ASoftware-Defined GPS and Galileo Receiver.A Single-Frequency Approach. Boston, MA:Birkhauser, 2007.
[59] B. M. Ledvina, S. P. Powell, P. M. Kintner,and M. L. Psiaki, BA 12-channel real-timeGPS L1 software receiver,[ in Proc. Nat. Tech.Meeting Inst. Navig., Anaheim, CA, Jan. 2003,pp. 767–782.
[60] G. W. Heckler and J. L. Garrison,BSIMD correlator library for GNSS softwarereceivers,[ GPS Solutions, vol. 10, no. 4,pp. 269–276, Nov. 2006.
[61] H. Hurskainen, J. Raasakka, T. Ahonen, andJ. Nurmi, BMulticore software-defined radioarchitecture for GNSS receiver signalprocessing,[ EURASIP J. Embedded Syst.,article ID 543720, 2009.
[62] T. E. Humphreys, J. A. Bhatti, T. Pany,B. M. Ledvina, and B. W. O’Hanlon,BExploiting multicore technology insoftware-defined GNSS receivers,[ inProc. 22nd Int. Meeting Sat. Div. Inst. Navig.,Savannah, GA, Sep. 2009, pp. 326–338.
[63] T. Hobiger, T. Gotoh, J. Amagai, Y. Koyama,and T. Kondo, BA GPU based real-time GPSsoftware receiver,[ GPS Solutions, vol. 14,no. 2, pp. 207–216, Mar. 2010.
[64] A. Mitelman, J. Almqvist, R. Hakanson,D. Karlsson, F. Lindstrom, T. Renstrom,C. Stahlberg, and J. Tidd, BTesting softwarereceivers,[ GPS World, vol. 20, no. 12,pp. 28–34, Dec. 2009.
[65] M. G. Petovello, C. O’Driscoll, G. Lachapelle,D. Borio, and H. Murtaza, BArchitectureand benefits of an advanced GNSSsoftware receiver,[ Positioning, vol. 1, no. 1,pp. 66–78, 2009.
[66] C. Fernandez-Prades, C. Aviles, L. Esteve,J. Arribas, and P. Closas, BDesign patternsfor GNSS software receivers,[ in Proc.5th ESA Workshop Satellite Navig. Technol.,Noordwijk, The Netherlands, Dec. 2010,pp. 1–8, DOI: 10.1109/NAVITEC.2010.5707981.
[67] X. Li and D. Akos, BImplementation andperformance of clock steering in a softwareGPS L1 single frequency receiver,[ Navig.,J. Inst. Navig., vol. 57, no. 1, pp. 69–85,Spring 2010.
[68] M. Anghileri, T. Pany, D. Sanroma-Guixens,J.-H. Won, A. Sicramaz-Ayaz, C. Stober,I. Kramer, D. Dotterbock, G. W. Hein, andB. Eissfeller, BPerformance evaluation of amulti-frequency GPS/Galileo/SBAS softwarereceiver,[ in Proc. 20th Int. Tech. Meeting Sat.Div. Inst. Navig., Fort Worth, TX, Sep. 2007,pp. 2749–2761.
[69] M. Fantino, A. Molino, and M. Nicola,BN-Gene GNSS receiver: Benefits ofsoftware radio in navigation,[ in Proc.Eur. Navig. Conf. Global Navig. SatelliteSyst., Naples, Italy, May 2009.
[70] B. W. Tolman, R. B. Harris, T. Gaussiran,D. Munton, J. Little, R. Mach, S. Nelsen,
B. Renfro, and D. Schlossberg, BThe GPStoolkitVOpen source GPS software,[ inProc. 17th Int. Tech. Meeting Sat. Div. Inst.Navig., Long Beach, CA, Sep. 21–24, 2004,pp. 2044–2053.
[71] D. Salazar, M. Hernandez-Pajares,J. M. Juan, and J. Sanz, BGNSS datamanagement and processing with theGPSTk,[ GPS Solutions, vol. 14, no. 3,pp. 293–299, Jun. 2010.
[72] M. A. Garcıa-Matatoros, D. Kuijper, andP. L. Righetti, BNAPEOS: ESA/ESOCnavigation package for earth observationsatellites,[ in Proc. Eur. Workshop Flight Dyn.Facilities, Darmstadt, Germany, Sep. 2003.
[73] J. M. Dow, R. E. Neilan, and C. Rizos,BThe International GNSS Service in achanging landscape of Global NavigationSatellite Systems,[ J. Geodesy, vol. 83, no. 7,pp. 191–198, Jul. 2009.
[74] L. L. Presti, X. Zhu, M. Fantino, andP. Mulassano, BGNSS signal acquisitionin the presence of sign transition,[ IEEEJ. Sel. Top. Signal Process., vol. 3, no. 4,pp. 557–570, Aug. 2009.
[75] D. Borio, L. Camoriano, and L. L. Presti,BImpact of GPS acquisition strategy ondecision probabilities,[ IEEE Trans. Aerosp.Electron. Syst., vol. 44, no. 3, pp. 996–1011,Jul. 2008.
[76] L. L. Presti, M. Fantino, and M. Nicola,BEnhanced Bayesian detection for weak GPSand Galileo signal acquisition,[ in Proc. Int.Tech. Meeting Inst. Navig., San Diego, CA,Jan. 2010, pp. 773–783.
[77] M. Fantino, P. Mulassano, F. Dovis, andL. L. Presti, BPerformance of the proposedGalileo CBOC modulation in heavymultipath environment,[ Wireless PersonalCommun., vol. 44, no. 3, pp. 323–339,Oct. 2007.
[78] C. Jekeli, Inertial Navigation Systems WithGeodetic Applications. Berlin, Germany:Walter de Gruyter, 2001.
[79] D. Titterton and J. Weston, StrapdownInertial Navigation Technology, 2nd ed.Reston, VA: IET, 2004.
[80] J. A. Farrell and M. Barth, The GlobalPositioning System & Inertial Navigation.New York: McGraw-Hill, 1999.
[81] H. Qi and J. B. Moore, BDirect Kalmanfiltering approach for GPS/INS integration,[IEEE Trans. Aerosp. Electron. Syst., vol. 38,no. 2, pp. 687–693, Apr. 2002.
[82] R. van der Merwe and E. A. Wan,BSigma-point Kalman filters for integratednavigation,[ in Proc. 60th Annu. MeetingInst. Navig., Dayton, OH, Jun. 2004,pp. 641–654.
[83] J. L. Crassidis, BSigma-Point Kalman filteringfor integrated GPS and inertial navigation,[in Proc. AIAA Guid. Navig. Control Conf.,San Francisco, CA, Aug. 2005, pp. 1–24.
[84] F. van Graas, A. Soloviev, M. U. de Haag, andS. Gunawardena, BClosed-loop sequentialsignal processing and open-loop batchprocessing approaches for GNSS receiverdesign,[ IEEE J. Sel. Top. Signal Process.,vol. 3, no. 4, pp. 571–586, Aug. 2009.
[85] C. Spagnol, R. Muradore, M. Assom,A. Beghi, and R. Frezza, BModel basedGPS/INS integration for high accuracy landvehicle applications,[ in Proc. IEEE Intell.Veh. Symp., Las Vegas, NV, Jun. 2005,pp. 400–405.
[86] A. Giremus, A. Doucet, V. Calmettes, andJ. Tourneret, BA Rao-Blackwellized particlefilter for INS/GPS integration,[ in Proc. IEEEInt. Conf. Acoust. Speech Signal Process.,
Montreal, QC, Canada, May 2004, vol. 3,pp. 964–967.
[87] M. Nguyen-H and C. Zhou, BImprovingGPS/INS integration through neuralnetworks,[ J. Telecommun., vol. 2, no. 2,pp. 1–6, May 2010.
[88] R. G. Brown and P. Y. C. Hwang,Introduction to Random Signal and AppliedKalman Filtering, 3rd ed. New York: Wiley,1997.
[89] M. G. Petovello, BReal-time integrationof a tactical-grade IMU and GPS forhigh-accuracy positioning and navigation,[Ph.D. dissertation, Dept. Geomatics Eng.,Univ. Calgary, Calgary, AB, Canada, 2003.
[90] N. Hjortsmarker, BExperimental system forvalidating GPS/INS integration algorithms,[Ph.D. dissertation, Dept. Comput. Sci. Elect.Eng., Lulea Univ. Technol., Lulea, Sweden,2005.
[91] G. Gao and G. Lachapelle, BA novelarchitecture for ultra-tight HSGPS-INSintegration,[ J. Global Position. Syst., vol. 7,no. 1, pp. 46–61, 2008.
[92] S. Alban, D. Akos, S. Rock, andD. Gebre-Egziabher, BPerformance analysisand architectures for INS-Aided GPStracking loops,[ in Proc. Inst. Navig. Nat.Tech. Meeting, Anaheim, CA, Jan. 2003,pp. 611–622.
[93] S. R. Babu, BUltra-tight integration ofGPS/pseudolites/INS: System design andperformance analysis,[ Ph.D. dissertation,Schl. Survey. Spatial Inf. Syst., The Univ.New South Wales, Sydney, N.S.W.,Australia, 2006.
[94] M. G. Petovello and G. Lachapelle,BComparison of vector-based softwarereceiver implementations with application toultra-tight GPS/INS integration,[ in Proc.19th Int. Tech. Meeting Sat. Div. Inst. Navig.,Fort Worth, TX, Sep. 2006, pp. 1790–1799.
[95] D. Li and J. Wang, BPerformance analysisof the ultra-tight GPS/INS integration basedon an improved Kalman filter design fortracking loops,[ in Proc. Int. Global Navig.Satellite Syst. Symp., Surfers Paradise,Australia, Jul. 2006, DOI: 10.1.1.71.6874.
[96] J. Wang and D. Li, BSystem designand performance analysis of extendedKalman filter-based ultra-tight GPS/INSintegration,[ in Proc. IEEE/ION PositionLocation Navig. Symp., San Diego, CA,Apr. 2006, pp. 291–299, DOI: 10.1109/PLANS.2006.1650615.
[97] M. Petovello, C. O’Driscoll, andG. Lachapelle, BWeak signal carrier trackingusing extended coherent integration withan ultra-tight GNSS/IMU receiver,[ inProc. Eur. Navig. Conf. Global Navig. SatelliteSyst., Toulouse, France, Apr. 2008, pp. 1–11.
[98] C. Fernandez-Prades, P. Closas, andJ. Vila-Valls, BNonlinear filtering forultra-tight GNSS/INS integration,[ inProc. IEEE Int. Conf. Commun., Cape Town,South Africa, May 2010, DOI: 10.1109/ICC.2010.5502037.
[99] B. Chan and M. Petovello, BCollaborativevector tracking of GNSS signals withultra-wideband augmentation in degradedsignal environments,[ in Proc. ION Int.Tech. Meeting, San Diego, CA, Jan. 2011,pp. 404–413.
[100] H. Niedermeier, B. Eissfeller, J. Winkel,T. Pany, B. Riedl, T. Worz, R. Schweikert,S. Lagrasta, G. Lopez-Risueno, andD. Jimenez-Banos, BDINGPOS: Highsensitivity GNSS platform for deep indoorscenarios,[ in Proc. Int. Conf. IndoorPosition. Indoor Navig., Zurich, Switzerland,
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
Vol. 99, No. 11, November 2011 | Proceedings of the IEEE 1903
Sep. 2010, DOI: 10.1109/IPIN.2010.5647542.
[101] T. Pany, J. Winkel, B. Riedl, M. Restle,T. Worz, R. Schweikert, H. Niedermeier,G. Ameres, B. Eissfeller, S. Lagrasta, andG. Lopez-Risueno, BPerformance of apartially-coherent ultra-tight coupledGNSS/INS pedestrian navigation systemenabling coherent integration times ofseveral seconds to track GNSS digitalsignals down to 1.5 dBHz,[ in Proc. 22nd Int.Meeting Sat. Div. Inst. Navig., Savannah, GA,Sep. 2009, pp. 919–934.
[102] L. De Hauwere, BDo you really meanbusiness in LBS?[ presented at the GalileoApplication Days, Brussels, Belgium,Mar. 2010, Consulting Group Ptolemus.
[103] A. Mitelman, P. Normark, M. Reidevall, andS. Strickland, BApples to apples:Standardized testing for high-sensitivityreceivers,[ GPS World, vol. 19, no. 1,pp. 16–33, Jan. 2008.
[104] F. D. Nunes, F. M. G. Sousa, andJ. M. N. Leitao, BPerformance analysisof altimeter-aided GNSS receiver forindoor scenarios,[ in Proc. 7th Conf.Telecommun., Santa Maria da Feria,Portugal, May 2009, pp. 1–4.
[105] V. Schwieger, BHigh-sensitivity GPS-Thelow cost future of GNSS?![ Int. Fed. Survey,Aug. 2007, pp. 1–15, Article of the Month.
[106] T. Pany, E. Gohler, M. Irsigler, andJ. Winkel, BOn the state-of-the-art ofreal-time GNSS signal acquisitionVA
comparison of time and frequencydomain methods,[ in Proc. Int. Conf. IndoorPosition. Indoor Navig., Zurich, Switzerland,Sep. 2010, DOI: 10.1109/IPIN.2010.5647537.
[107] T. Pany, B. Riedl, J. Winkel, T. Worz,R. Schweikert, H. Niedermeier, S. Lagrasta,G. Lopez-Risueno, and D. Jimenez-Banos,BCoherent integration time: The longer, thebetter,[ Inside GNSS, vol. 4, no. 6, pp. 52–61,Nov./Dec. 2009.
[108] M. G. Petovello, C. O’Driscoll, andG. Lachapelle, BUltra-tight GPS/INS forcarrier phase positioning in weak-signalenvironments,[ in Proc. NATO RTO-SET-104Symp. Military Capabilities Enabled AdvancesNav. Sensors, Antalya, Turkey, Oct. 2007,pp. 1–18.
[109] J. A. Del Peral-Rosado, J. A. Lopez-Salcedo,G. Seco-Granados, J. M. Lopez-Almansa, andJ. Cosmen, BKalman filter-based architecturefor robust and high-sensitivity tracking inGNSS receivers,[ in Proc. 5th ESA WorkshopSatellite Navig. Technol., Noordwijk,The Netherlands, Dec. 2010, DOI: 10.1109/NAVITEC.2010.5708005.
[110] R. Scopigno and H. Cozzetti, BMobile slottedaloha for VANETs,[ in Proc. IEEE 70thVeh. Technol. Conf., Anchorage, AK,Sep. 2009, DOI: 10.1109/VETECF.2009.5378792.
[111] R. Scopigno and H. Cozzetti, BGNSSsynchronization in VANETs,[ in Proc.3rd Int. Conf. New Technol. Mobility Security,
Cairo, Egypt, Dec. 2009, DOI: 10.1109/NTMS.2009.5384821.
[112] D. Margaria, L. L. Presti, N. Kassabian, andJ. Samson, BA new peer-to-peer aidedacquisition approach exploiting C=N0
aiding,[ in Proc. 5th ESA WorkshopSatellite Navig. Technol., Noordwijk,The Netherlands, Dec. 2010, DOI: 10.1109/NAVITEC.2010.5708021.
[113] S. Digenti, M. Nicola, L. L. Presti, andM. Pini, BTechnique for the estimation ofPC clock offset in a GNSS-aided networkof collaborative users,[ in Proc. 5th ESAWorkshop Satellite Navig. Technol.,Noordwijk, The Netherlands, Dec. 2010,DOI: 10.1109/NAVITEC.2010.5707990.
[114] M. Rao, L. L. Presti, and J. Samson, BPeer topeer equation augmentation for an altitudeaided GNSS receiver,[ in Proc. IEEE 72ndVeh. Technol. Conf., Ottawa, ON, Canada,Sep. 2010, DOI: 10.1109/VETECF.2010.5594103.
[115] L. L. Presti, D. Margaria, and M. Rao, BNoveltechniques for a cooperative positioningapproach based on peer-to-peer networks,[in Proc. Int. Conf. Data Flow From Space toEarth, Venezia, Italy, Mar. 2011, pp. 1–8.
[116] C. J. Hegarty and E. Chatre, BEvolutionof the Global Navigation Satellite System(GNSS),[ Proc. IEEE, vol. 96, no. 12,pp. 1902–1917, Dec. 2008.
ABOUT THE AUT HORS
Carles Fernandez-Prades (Member, IEEE) re-
ceived the M.Sc. and Ph.D. (cum laude) degrees
in electrical engineering from the Universitat
Politecnica de Catalunya (UPC), Barcelona, Spain,
in 2001 and 2006, respectively.
During the pursuit of his Ph.D., he was the
recipient of a 2001–2004 Ph.D. scholarship
granted by the Spanish Ministry of Education. He
joined the Department of Signal Theory and
Communication, UPC, as a Research Assistant,
getting involved in a number of European, Spanish, and Catalan research
projects, as well as industrial ones. In 2002, he engaged in an educational
innovation project funded by the Catalan Government. In 2003, he
became the Technical Manager at UPC for a European Space Agency
research project. He was Teaching Assistant in the field of Analog and
Digital Communications at UPC (2001–2005), and he directed six M.S.
theses. In May 2006, he joined the Centre Tecnologic de Telecomunica-
cions de Catalunya (CTTC), where he currently holds a position of a
Research Associate and the Coordinator of the Communications Sub-
systems Area. His primary areas of interest include statistical signal
processing, estimation theory, GNSS synchronization, digital commu-
nications, and design of radio-frequency (RF) front-ends. During the
second semester of 2007, he was a Visiting Lecturer at the Universidad
Tecnologica Metropolitana (UTEM), Santiago, Chile.
Letizia Lo Presti (Member, IEEE) was born in
Palermo, Italy, on December 13, 1947. She gradu-
ated in electronic engineering from the Politecni-
co of Torino, Torino, Italy, in 1971.
Currently, she is a Full Professor with the
Information Engineering Faculty, Politecnico di
Torino, working in the Electronics Department.
She is the head of the NavSAS research group. Her
research activities cover the field of digital signal
processing, simulation of telecommunication sys-
tems, and the technology of navigation and positioning systems. Her
teaching activity is mainly focused on signal processing (from the
fundamentals to advanced concepts, such as array processing, statistical
signal analysis, time-frequency distribution, and estimation theory),
digital communications, and algorithms for GPS and Galileo receivers.
She is the scientific coordinator of the Master on Navigation and Related
Applications held by Politecnico di Torino (since 2003). She actively
cooperates with the officers of the United NationsVOffice for Outer
Space Affair (UN-OOSA), Vienna, Austria, in the framework of the UN/
Italy fellowship program, with the aim of keeping the Master program
permanently aligned with the UN needs and suggestions.
She is a member of the Working Group of the International Committee
of Global Navigation Satellite Systems (ICG) led by the UN-OOSA.
Emanuela Falletti received the M.Sc. and Ph.D.
degrees in electronics and communications engi-
neering from Politecnico di Torino, Torino, Italy, in
1999 and 2004, respectively.
Currently, she is with the Istituto Superiore
Mario Boella, Torino, Italy, where she is responsi-
ble for projects on the analysis and design of
signal processing algorithms for GNSS digital
receivers. Her research interests have focused on
array signal processing, wireless channel model-
ing, communications from high altitude platforms, and signal processing
for digital receivers.
Fernandez-Prades et al. : Satellite Radiolocalization From GPS to GNSS and Beyond
1904 Proceedings of the IEEE | Vol. 99, No. 11, November 2011