Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

COMBINED URBAN AND INDOOR NETWORK PLANNING

USING THE DOMINANT PATH PROPAGATION MODEL


René Wahl, Gerd Wölfle

AWE Communications GmbH, Otto-Lilienthal-Str. 36, 71034 Boeblingen, Germany, www.awe-communications.com

ABSTRACT 2. TYPES OF DATABASES


With the growing interest for broadband mobile The structure of a scenario to be predicted is stored
services in mobile communication networks, the in a database file. Depending on the type of scenario
investigation of radio transmission in and into buildings (indoor, urban or rural), a different description of the
is getting more important. Popular empirical and relevant obstacles is necessary.
deterministic models for the propagation inside
buildings compute the field strength based on the inner
structure of the buildings (walls, furniture). But for
current and future wireless networks (3G, B3G, W-
LAN, WiMax,..), the neighboring buildings must also
be considered to avoid interference problems in these
buildings. Additionally the indoor coverage of outdoor
transmitters must be analyzed to guarantee a high QoS
even inside buildings. A new concept for the prediction
of the field strength in such hybrid scenarios (urban and
indoor) is presented in this paper. This new concept
does not rely only on the direct ray (like empirical
models) and it does not consider hundreds of rays for a Figure 1. Vector building data of a part of a city (with
single radio link (like ray tracing). The new model pixel based DTM data).
focuses on the most dominant path between transmitter For urban scenarios, databases usually have limited
and receiver. It allows the computation of the transition 3D information (sometimes called 2.5D), i.e. for each
from an urban to an indoor scenario and vice versa, thus building a polygonal cylinder with its ground plane
allowing an accurate computation of the received power (polygon) and its uniform height is defined (see figure
inside and around buildings. The new model is validated 1). The material properties of each building can be
with various measurements and shows a good accuracy. defined individually (or default properties for all
buildings can be used if no detailed information is
1. INTRODUCTION available).
After the great success of wireless communications
used in land and personal mobile radio networks, the
wireless communication inside buildings becomes more
and more important. Not only W-LAN and WiMax, but
also 3G and B3G networks with their wireless
multimedia services (e.g. video terminals) are used
inside and outside buildings and rely on the same high
Quality of Service (QoS) everywhere. To guarantee this
QoS, the operator has to analyze the interaction between
indoor and outdoor wave propagation during the
network planning process. Unfortunately the current
propagation models are always dedicated to a single Figure 2. Complex vector oriented 3D database of a
scenario (either urban or indoor). Interactions (e.g. large multi floor building.
interference) between indoor and outdoor transmitters
cannot be computed with the required high accuracy. In contrast to the urban case, indoor databases
Additionally the indoor coverage of outdoor transmitters consist of polygons with arbitrary shapes and
cannot be computed based on indoor as well as outdoor orientations. The highly accurate definition of an
building data. In the following sections, a new arbitrary number of objects with different materials is
propagation model as well as the approach to compute possible. Additionally subdivisions with different
urban/indoor transitions will be described. material properties (e.g. doors or windows) can be

_____________________________________________________
Proc. ‘EuCAP 2006’, Nice, France
6–10 November 2006 (ESA SP-626, October 2006)
modeled. Fig. 2 shows an example of an indoor ray between transmitter and receiver. In urban scenarios
database of a large multi floor building. this ray can include the over-rooftop diffractions (e.g.
For the penetration from an urban scenario into an Walfisch-Ikegami model) while in indoor scenarios the
indoor environment, or vice versa, the databases of both ray is always the direct line between Tx and Rx. In both
scenarios – indoor and urban – have to be combined. scenarios, this direct ray is not always dominant as it is
The objects in the indoor database are significantly highly attenuated. A model based on this path must lead
smaller than the urban buildings. If the same to errors in scenarios where the direct ray is contributing
discretization would be used for both (urban and indoor) only a very small part to the total received signal power.
objects, this would lead to memory requirements which
could not be handled with standard PCs. As already
mentioned, for the urban propagation model a limitation
of the objects to horizontal (walls) and vertical (roofs)
elements of larger size accelerates the prediction T T T
without reducing the accuracy significantly. In contrast
to this, the indoor propagation model has to consider R R R
arbitrarily oriented elements which increase the Figure 4. Empirical models use the direct path (left),
computation time. This emphasizes the need for two Ray Tracing uses many paths (middle) and DPM uses
different types of databases (urban and indoor) and the the most relevant path (right).
requirement for a smooth transition between the urban
and the indoor models. In the picture in the center of fig. 4 the principle of
ray-optical propagation models is shown. Up to
hundreds of rays can be computed for each receiver.
The contributions of all rays are superposed (in most
cases incoherent superposition) to obtain the received
power. But in most cases only 2 or 3 rays are
contributing more than 98% of the energy, i.e. by
focusing on these dominant rays the accuracy would be
sufficient [3].
Figure 3. Shape of indoor building in an urban A second disadvantage of ray-optical models is their
scenario. high dependency on small inaccuracies in the databases.
As angular criteria are evaluated during the ray-optical
The interface between the two types of databases is
prediction, the orientation of walls is extremely
automatically determined. A polygonal cylinder with
important. Unfortunately databases with this very high
vertical walls and a flat roof is put around the indoor
accuracy are not easily available for most applications.
building database. So all walls of the indoor database
In addition to the two disadvantages mentioned above,
are inside the polygonal cylinder (see fig. 3).
another problem of ray-optical models arises: Either the
The material properties of the cylinder can be
computation time is very long or, if a preprocessing is
defined and are used in the urban database to determine
done, the computation time for the preprocessing is high
the reflections and diffractions with the urban
(During the preprocessing the database is divided into
propagation models. While the urban propagation model
tiles and segments and the visibility between these tiles
considers only the cylinder, the indoor propagation
and segments is determined. As databases consist of a
model ignores the cylinder and considers only the walls
large number of tiles and segments, this process can
(objects) of the indoor database inside the cylinder. So
take a considerable amount of time – from some hours
the cylinder is transparent for the indoor models (rays
up to several days for very large scenarios [1]).
are passing the cylinder without any attenuation) and the
cylinder is not considered during the computation of the
3.2. Requirements for a new Model
indoor prediction.
The resolution of the result matrix is also adapted to After analyzing the status of the models currently
the scenario type. Inside the cylinder, a higher available, the requirements for a new model can be
resolution is used, whereas in the urban domain the defined:
resolution of the matrix is lower. Reasonable values are ƒ Focusing on the dominant paths and not
0.5 m for indoor environments and 5 m for urban areas. computing hundreds of paths with small
contributions.
3. DOMINANT PATH MODEL ƒ Model should not depend on each micro-detail in
the vector database.
3.1. Classical Prediction Models
ƒ Simple auto-calibration possible with reference
The left picture in fig. 4 shows the problem of data (e.g. measurements).
empirical propagation models. They rely on the direct
ƒ The duration of preprocessing of the database of wall number j, i.e. t2 is the transmission loss of the
has to be short. second wall penetrated by the dominant path. In pure
urban scenarios transmissions are not considered.
3.3. The Dominant Path Model The algorithm of the Dominant Path Model is
suitable for all types of scenarios: Rural, urban or indoor
Based on the requirements defined in the previous
environments require only slight adaptations of the
subsection, the algorithm of the Dominant Path Model
algorithm.
(DPM) can be subdivided into two steps:
ƒ Determination of the dominant paths
1) Sub-model for urban scenarios
(geometry)
Predictions in urban scenarios are accomplished
ƒ Prediction of the path loss along the paths.
with the Urban Dominant Path Model. The
Determining the dominant paths is not a very simple
determination of the paths and the prediction of the
task. The algorithm and its performance are published in
signal level is done as explained in the section before.
[2].
The prediction of the path loss along the path is done
by using the following equation:
⎛ 4π ⎞ n m
L = 20 log⎜ ⎟ + 10 p log(l ) + ∑ f (ϕ , i ) + ∑ t j − Ω
⎝ λ ⎠ i =1 j =1
L is the path loss in dB of a path with a length l (in
meters). λ is the wavelength and Ω the waveguiding
factor (see below). f(φ,i) is a function which determines
the interaction loss in dB, i.e. the loss when changing
the direction of propagation. The angle between the
former direction and the new direction of propagation is
φ. The loss increases linearly with the angle, starting
with an offset α1. The linearity ends at angle φ1 and the
Figure 5. Prediction of field strength in a hilly city. Five
loss will be constant (at level α2) for angles above φ 1. i
typical dominant paths are shown.
is the number of the interaction, i. e. i=2 means the
second interaction on this propagation path. The number In hilly scenarios, the topography must be
of the current interaction is important, because not all considered additionally to the buildings as it may
interactions are weighted in the same way. Interactions influence the visibility. Fig. 5 shows a prediction of a
with higher indices i lead to reduced losses compared to part of a city with a hilly topography in the downtown
interactions with smaller indices i. This is considered area. The area in the upper right part shows lower
because the wave is becoming more and more diffuse received power values due to the shadowing effect of
after each interaction and the more options for the topography. In this area the results would be too
interactions occur the more diffuse the wave will be optimistic, if topography was not considered. Thus, it is
(more options means less total loss). Losses due to evided to consider terrain data in the UDP Model.
horizontal and vertical changes in the direction of
propagation are determined independently.
The factor p depends on the visibility between the
current pixel and the transmitter. Adapting p to the
situation allows different path loss exponents and
individual breakpoints depending on the LOS (Line of
Sight) or NLOS (Non Line of Sight) condition.
The waveguiding factor Ω is described in [2]. The
reflection loss of the walls along the path, as well as
their distance to the path, influence α. The smaller the
reflection loss and the closer the wall to the path, the
higher the waveguiding factor. The gain due to
waveguiding is determined for each pixel before the
prediction starts. During the prediction this gain is Figure 6. Gain due to waveguiding effects in an urban
accumulated along the propagation path. In close street scenario.
canyons or in small corridors this gain is higher than in
open areas. As mentioned above, a waveguiding factor can be
For the determination of the path loss in indoor considered during computation (see fig. 6) to increase
scenarios, transmissions through walls have to be the accuracy. The additional computation time for the
considered additionally. tj means the transmission loss determination of the waveguiding effects for the
scenario in fig. 6 is less than one second, so this should an urban to an indoor scenario or vice versa. Therefore
always be included. both prediction models are used during the computation,
The algorithm works in rigorous 3D mode, so the depending on the current domain.
wave can propagate over all buildings and obstacles and In section 2, the interface between urban and indoor
is not limited to the horizontal plane (this over-rooftop databases was described. When passing the shape of an
diffraction can be disabled optionally for higher indoor building (that means a transition from urban to
frequencies or very low antenna heights). indoor domain or vice versa), the actual computed
propagation path is passed to the corresponding
2) Sub-model for indoor scenarios prediction sub-model, which then further processes the
In case of an indoor scenario the Indoor Dominant computation of this ray.
Path Model (IDP) is used for the prediction. In addition For indoor and outdoor environment, different
to the effects taken into account in urban computation parameters for the DPM can be defined (path loss
mode, the transmission loss of walls penetrated by the exponents, interaction losses, etc.) to obtain highest
dominant path must be considered as well. accuracy in indoor as well as in outdoor mode. During
the transition from one scenario to the other, the
according set of parameters (outdoor/indoor) is selected
and used for the further prediction.
The prediction in the urban area is done on only one
height level, whereas in the indoor environment
multiple prediction layers can be defined and are
considered during the computation.

4. EXAMPLES
4.1. Transmitter located in urban scenario
Figure 7. Prediction of received power with multiple In fig. 9 a combined urban/indoor prediction of a
layers. part of Hong Kong is shown. The transmitter
(isotropical radiator, 40 dBm, 1.8 GHz) is located on a
The path loss exponent p is adapted to indoor hill. In the scenario, the indoor walls of two buildings
scenarios. Three different states are defined: LOS (Line are available. For predictions inside these buildings, the
of Sight), OLOS (obstructed LOS) and NLOS (non- UDP passes the actual propagation path to the IDP,
LOS). For each state individual path loss exponents can which continues with the prediction inside the building,
be defined. taking into account all indoor walls. A hand over back
For buildings with multiple floors, the IDP is able to to the UDP when leaving the indoor scenario is of
compute the dominant paths on several layers course also possible."
simultaneously (see fig. 7). This 3D mode leads to
highly accurate results even in environments with more
than one floor. Indoor Transmitter
Building 1

Indoor Building 2

Figure 8. Gain due to waveguiding effects in a building. Figure 9. Prediction of a combined urban/indoor
scenario with two indoor buildings.
As already known from the urban domain,
waveguiding effects can be considered (fig. 8). In For the prediction a resolution of 5 m is used in the
indoor scenarios this is implemented in the same way. urban domain and 0.5 m for the indoor environment.
Inside the buildings, the path loss is predicted on 4
3) Combination of indoor and urban sub-model layers. Fig. 10 shows the results of the prediction for
The new approach presented in this paper allows a one of the 4 layers inside building 1.
combination of UDP and IDP for the penetration from
The computation time for the whole scenario (urban The predicted area is about 4.0 km x 4.3 km. Resolution
domain and 4 indoor layers for each building) is not is 10 m. As shown in the table, the results of the
longer than 5 minutes. The prediction area is 2.6 km x predictions are quite good.
3.0 km.

Tx

Figure 10. Predicted distribution of field strength on


Figure 12. Stuttgart database with topography.
third floor of building 1.

4.2. Transmitter located inside building


The transmitter (isotropical radiator, 20 dBm, 3.5
GHz) is now located inside a building in a small city.
The prediction is again accomplished in urban (5 m
resolution) and indoor (0.5 m) domain. Inside the
building, 4 layers are predicted. The left picture in fig. Figure 13. Prediction for transmitter location 1 (left)
11 shows the results in the urban domain, taking into and difference from prediction to measurement route.
account the indoor walls and their transmission losses.
In the right picture, the distribution of the field strength Table I. Comparison between predictions and
in the indoor environment for the 10th floor, where the measurements for all measurement routes in Stuttgart.
transmitter is located, is depicted. Tx Mean Std. Comp.
Location Value Deviation Time *
1 1.63 dB 6.12 dB 178 s
2 -0.55 dB 6.33 dB 115 s
3 4.84 dB 6.30 dB 68 s
* Standard PC with 1.8 GHz CPU and 2 GB RAM

5.2. Indoor Comparison


Figure 11. Prediction result for urban scenario (left) The DPM algorithm is also validated in indoor
and indoor predicion inside building 1 on third floor. environments. Therefore a scenario in a modern
multifloor office building is chosen. The database is
shown in figure 14.
5. COMPARISON TO MEASUREMENTS
5.1. Urban Comparison
For the validation of the Dominant Path Model in
urban scenarios, a comparison of predicted results to
measurements is made. Therefore a measurement
campaign in the downtown area of Stuttgart (Germany)
is accomplished (see fig 12).
The transmitter (sector antenna, 34 dBm EIRP, 3.5
GHz) is located on three buildings with different
heights. Each route is more than 10 km in length. The
prediction and the differences between predicted and Figure 14. The office building database.
measured values for route 1 are shown in figure 13. A Measurements for several transmitter locations are
statistical evaluation of the differences between available [3]. Each time, the transmitter (isotropical
measurements and predictions for both routes are listed radiator, 20 dBm at 1800 MHz) is located in a different
in table I. place. Figure 15 shows a prediction result as well as the
difference from prediction to measurements for indoor scenario. In the right picture, the difference
transmitter location 12. between predictions and measurements is depicted. A
statistical evaluation for both transmitter locations 1 and
2 is presented in table III. Again, the results are
promising.

Figure 15. Prediction for transmitter location 12 (left)


and difference from prediction to measurement route Figure 17. Prediction for transmitter location 1 (left)
(right). and difference from prediction to measurement route
The statistical evaluation for all investigated (right).
measurement routes and the computation times are Table III. Comparison between predictions and
presented in table II. The results are again quite good. measurements for both measurement routes in the
combined urban/indoor scenario.
Table II. Comparison between predictions and
measurements for all measurement routes in the multi Tx Mean Std. Comp.
floor building. Location Value Deviation Time *
1 0.87 dB 5.52 dB 18 s
Tx Mean Std. Comp. 2 3.54 dB 6.55 dB 18 s
Location Value Deviation Time *
1 -3.52 dB 4.71 dB 4s 6. CONCLUSIONS
5 -0.46 dB 3.67 dB 4s
9 0.65 dB 5.75 dB 4s In this paper a new propagation model for the
11 2.38 dB 6.02 dB 3s transition between urban and indoor scenarios is
12 0.16 dB 4.34 dB 4s presented. The model is based on the dominant paths
and was adapted to handle objects with arbitrarily
located and rotated planes (typical for indoor) as well as
5.3. Combined urban/indoor Comparison
very large areas (urban cells with cell radii of several
In order to validate the approach in a combined kilometers). A transition interface is described for the
urban/indoor scenario, a measurement campaign using computation of the indoor penetration (coverage) as
CW signals was performed. Especially for building well as for the outdoor interference. Measurement
penetration effects a measurement setup was chosen to campaigns confirmed that the model delivers very
obtain measurements directly in front of the building accurate results.
walls and behind the walls. A building with surrounding
balcony was therefore selected (see figure 16). Details 7. REFERENCES
of the measurement setup and the measurement results
[1] R. HOPPE, P. WERTZ, F. M. LANDSTORFER and G.
are published in [4]. The transmitter (isotropical
WÖLFLE, “Advanced ray-optical wave propagation
radiator, 40 dBm, 1500 MHz) was located outside on a
modelling for urban and indoor scenarios“, Euro.
neighboring building.
Trans. Telecomms. 2003.
[2] G. WÖLFLE and F. M. LANDSTORFER, “Dominant
Path for the Field Strength Prediction“, 48th IEEE
VTC 1998, Ottawa (Canada).
[3] G. WÖLFLE, “Adaptive Propagation Models for the
Planning of Wireless Communication Networks and
for the Computation of the Reception Quality inside
Buildings”, PhD thesis, University of Stuttgart,
2000.
Figure 16. The combined scenario database with the [4] R. HOPPE, G. WÖLFLE, and F. M. LANDSTORFER,
indoor building and the two transmitters (blue balls). “Measurement of Building Penetration Loss and
Propagation Models for Radio Transmission into
The left picture in fig. 17 shows the result of the Buildings “, 50th VTC 1999, Amsterdam (The
predicted field strength for Tx 1 using the Dominant Netherlands).
Path Model with the transition from the urban to the

You might also like