Computational Intelligence - February 2014 - 45

omnidirectional antenna, but few studies
have been conducted on the systems
equipped with directive antenna.
In this paper, we will propose a novel
WCDMA network planning model
based on iterative power control scheme
and optimize it by utilizing a multiobjective evolutionary algorithm based on
decomposition. The main contributions
of this paper are four-fold:
1) Propose a simple, but effective
model: We pre-determine the maximum influence region of each BSs'
sector. Then, the total interference is
approximated with the interference
due to the signals transmitted by the
emitter (i.e., mobile stations in the
uplink and base stations in the
downlink) in the same maximum
influence reg ion. This model
neglects the interference due to the
channels without maximum influence region, thus simplifying the
WCDMA planning model without
degrading the performance evidently.
2) Develop an iterative power control
scheme for directive antenna: In this
scheme, when a single transmitted
power is changed, we only update
the transmitted power of the active
connections in the same cell. Such
power control method significantly
reduces the complexity of the model.
3) Sample a small, but representative set
of combination levels of configuration parameters: In the proposed
model, each BS is equipped with
directive antenna and four configuration parameters (i.e., antenna
height, antenna tilt, sector orientation, and pilot signal power) are considered. However, since there are too
many combination levels of configuration parameters, it is very difficult
to find the best one from them. To
overcome this problem, a representation method based on orthogonal
design is proposed in this paper.
4) Apply a novel multiobjective evolutionary algorithm to solve this
model: The evolutionary algorithm,
as one of the most powerful tool to
solve the complex optimization
problems, has been widely used in
the various fields [23]-[26]. Accord-

ingly, we present a multiobjective
evolutionary algorithm based on
decomposition [27] to solve this
combinatorial optimization problem.
Simulation results have shown the
effectiveness of the proposed algorithm by providing a set of high
quality solutions.
The remainder of this paper is organized as follows: Section II proposes the
mathematical model based on the local
interference for network planning problem and describes the iterative power
control scheme. In Section III, we code
the solution with a novel representation
strategy based on orthogonal design and
present a multiobjective evolutionary
algorithm to solve this model. Computational results are reported in Section IV.
Finally, a conclusion is drawn in Section V.
II. The Model Based on Local
Interference and Iterative
Power Control Scheme

In WCDMA network planning problem, it needs to select a subset of candidate sites (CSs) to install BSs with suitable configuration parameters and assign
the test points (TPs) to an available BS.
In this section, we will present a model
for the WCDMA network planning.
Accordingly, a set of TPs and a set of
CSs are given as follows:
TP: The service area is divided into
some square grids. As shown in Fig. 1,
the centroid of a grid is regarded as a TP.
The ith TP is denoted as TP i with
i ! I = {1, 2, 3, gm}, where m is the
number of TPs. Moreover, the traffic
demand of TP i is represented by u ui in
the uplink and u di in the downlink. It
can simply correspond to the number of
the active connections. Then the total
traffic demand of TP i is u i = u iu + u di .
CS: Suppose there are n CSs and
there is at most one BS to be installed at
each CS. For convenience, we will not
discriminate between the base station
installed at CS and the Candidate Site.
For each BS, four configuration parameters are considered in this paper. That is:
❏ Antenna height: h b
❏ Antenna Tilt: b
❏ Sector Orientation: c
❏ Pilot Power: pt .

Low

Heavy

cSector1
Sector3
BS2

BS1
Sector2
TPi1 BS
3

Figure 1 Illustration of the no-load coverage, low-load coverage, heavy-load coverage
and sector orientation.

Then, a set of combination levels, i.e.
K = {1, 2, g, l}, of the configuration
parameters is given for each BS, where l
is the number of the combination levels
of configuration parameters. Moreover,
an installation cost c jk is also given to
each base station installed at CS
j ! J = {1, 2, f, n} with combination
level k ! K of configuration parameters.
Obviously, the installation cost varies
with the BS's configuration parameters.
A. The Impact of the Configuration
Parameters on Capacity

The first two configuration parameters,
antenna height and antenna tilt, have an
impact on propagation gain tensor. The
sector orientation defines the sets of TPs
within the same sector and the pilot
power is utilized to determine the BS, to
which each TP is assigned. The impact
of these configuration parameters will
be described as follows:
The antenna height and tilt have an
impact on the propagation gain tensor
G = [g ijk], where 0 < g ijk < 1 is the
propagation gain from TP i to BS j
with the combination level k of configuration parameters. It can be estimated by using prediction tools or
obtained by actual measurements. We
use the COST-231 Hata model [29]
and vertical diagram [28] to describe
the impact of antenna height and tilt
on the propagation gain in this paper.
It can be represented by the attenuation in db, i.e. g ijk = 1/10 (lijk /10), where
l ijk is the propagation factor between
TP i and BS j with configuration

February 2014 | Ieee ComputatIonal IntellIgenCe magazIne

Table of Contents for the Digital Edition of Computational Intelligence - February 2014