IEEE Robotics & Automation Magazine - March 2014 - 58

In strongly constrained environments, the propagation phenomena surrounding the antenna, more precisely
diffraction and multipath, are responsible for severe errors in
the raw observables (pseudoranges and-to a lesser extent-
Doppler measurements) that are computed by receivers. Diffraction is a usual error
source in GPS positioning
GNSSs have significant
if obstruction exists and
occurs whenever the
potential in the
direct signal is obstructed,
but a diffracted signal is
development of intelligent
received and processed.
However, the most severe
transport systems (ITSs) and deviations (up to several
tens of meters) occur with
associated services.
multipath, especially
when the reflected path is
only tracked and the
direct one is blocked. These signals, coming from invisible
satellites, are called NLOS signals and are relatively frequent
in cities. A recent study carried out in the Nantes city center,
with 15-m-high buildings and narrow streets (10-m maximum), has shown a rate of NLOS signals, among all the measurements output by the receiver, between 20 and 25% for a
standard high-sensitivity GPS receiver [2].
This article describes a new technique for improving the
QoS of GNSS positioning in deep urban environments
through the detection of the NLOS signals by using digital
maps describing the 3-D environment. The scope of this article will be limited to the map-aided NLOS detection principle
and methods; the usage of this detection into PVT computation algorithms will be addressed in further articles.
Related Articles
It goes beyond the scope of this article to detail the various
techniques that were designed by the manufacturers of GNSS
receivers for multipath mitigation. We will just mention that
they generally address either antenna or correlator technologies. For further information, a summary can be found in [3].
Most receivers, even mass-market original equipment manufacturer ones, already have embedded multipath rejection
tracking loops. Antennas also benefit from polarity-selective
gain patterns, notwithstanding that low-cost patch antennas
still remain highly sensitive to any signal, whether it is direct
or reflected.
Navigation algorithms as well as GNSS hardware design
are important for multipath mitigation. Once pseudoranges
have been made available for position and time computation,
failure detection is the first integrity test level: the normalized
sum of the squared residuals is compared with a | 2 threshold
with N - 4 degrees of freedom (with N being the number of
satellites in view) in association to a given probability of false
alarm. At a more advanced integrity test level, fault detection
and exclusion (FDE) consists in identifying one faulty measurement in the redundant data set. It has to be underlined
that a unique default assumption is made and that redun58

IEEE ROBOTICS & AUTOMATION MAGAZINE

MARCH 2014

dancy is needed. However, none of the outlined techniques
works well in cities, where satellite availability is low and
where multipath can cause several faulty measurements at the
same time. The authors of [4] suggest an alternative approach
to the standard FDE that cannot work under a limited number of satellites, consisting in a characterization of the solution
error (based on residuals) and the provision of a protection
level with all satellites in view. This protection level is generally larger than if exclusion was done, which leads to reduced
application availability. However, at least five satellites are
required to figure out positioning system residuals.
For a few years, in addition to navigation process improvements, researchers have addressed the use of 3-D models of
the environment to analyze reception conditions and mitigate
multipath phenomena.
The LocoPROL project [5] has already used an obstacle
elevation model from both sides of a railway acquired by video
cameras. In [6], a fish-eye infrared camera is used to map satellite positions with respect to the surrounding buildings. More
recently, the CAPLOC project [7] addresses this issue for
guided transport in urban environments, using a fish-eye camera on the train to build a 3-D model from the successive
images in real-time.
Our approach relies on using 3-D models without extra
video sensor and image processing techniques, i.e., digital
maps that can be used in real time by an on-board equipment.
Two different maps with two different ways to use them will
be presented and discussed in this article: the first map is built
by hand from Google Earth while the second one is much
more industrial since it is a commercial product from the
French survey institute IGN, accessible through a dedicated
geographical information system (GIS).
Method 1: Elevation-Enhanced
Map Using Google Earth
EEmap Creation
The fundamental paradigm of the elevation-enhanced map
(EEmap) proposal is to include only the most basic information needed to detect whether or not a facade blocks the
direct view of a GNSS satellite from the receiver antenna.
For that purpose, the description of the buildings follows
the format given in Table 1, where subscripts 1 and 2 stand
for the two upper corners of the facade under consideration, and Lat, Lon, w, and H stand for, the latitude, the
longitude of the corners, the width, and the height of the
building, respectively.
The process of creating an EEmap is essentially manual
and it is based on the exploitation of Google Earth images.
The first step to create the EEmap is the extraction of the geodetic coordinates of the two upper corners of the facade of the
Table 1. Building model parameters.
Bldg ID

Lat 1

Lon 1

Lat 2

Lon 2

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2014