Computational Intelligence - February 2017 - 30

algorithmic efficiency can be greatly improved. A new coding scheme inspired by the famous four-color theorem is
specially designed for this multi-objective TA planning
model. Computer simulations are conducted and the quality
of the new model is confirmed by comparing the results of
the multi-objective model with those of a single-objective
model. The essential role of the population decomposition
strategy has also been identified by comparing the proposed
algorithm with the Multi-objective Evolutionary Algorithm
based on Decomposition (MOEA/D).

I. Introduction

ith the development of mobile communication
networks, the LTE network has become ever
more popular around the world. The study of
LTE networks has become a hot issue in the theory and practice of contemporary mobile communication networks [1]. Location management is an essential task in LTE
networks, and it can directly affect the stability, security and
performance of the networks. Location management in LTE
networks aims at quickly tracking where the users are, and this
tracking makes it possible to deliver calls, short message services
and other mobile phone services to the users in a timely way.
In the management of an LTE network, cells are bound
together to form a series of TAs, and then the TAs are further
grouped into TA lists (TALs). The main function of TALs is to
track the locations of a user equipment (UE). Each TAL has an
identifier known as its Tracking Area Identity (TAI), which is
used for the location update of UEs. All the Base Stations (BSs)
in the same TAL broadcast the same TAI regularly through a
broadcast control channel. UEs can recognize the TAI and store
it in the subscriber identity module (SIM) when registering
with the network. If the registered TAI of a UE is found different from the current broadcast TAI, location update is triggered.
Thus, when a user enters a different TAL, the UE's location
needs to be updated. Obviously, the more the TAL boundary
crossings is, the more location updates the network performs. In
the process of location update, UE updates its location and notifies its current location to the network [2]. When there is a
phone call for a UE, the network will search for this UE. This
search is known as paging. The most simple and intuitive way of
paging is to check each cell one by one, which is called the
blanket polling paging. This registered TAL information can narrow the search into a certain TAL, because only the cells belonging to the TAL where the UE is registered need to be paged.
Paging and location update lead to two different kinds of
costs which are termed as location management cost. If we
enlarge the TAL to the extreme situation, namely, making all
the cells into one TAL, we can eliminate the location update
cost completely. However, the large size of the TAL leads to the
need to search more cells to ensure a successful paging, which
means more resources need to be expended. In addition, the
load of every single cell increases because of frequent paging.
In the limit, the paging success rate decreases so that the entire
network becomes unstable and its service quality cannot be

IEEE ComputatIonal IntEllIgEnCE magazInE | FEbruary 2017

guaranteed. At the other extreme, by making each cell an independent TAL, we can minimize the paging cost of the entire
network, but we also maximize the location update cost. In
fact, the two objectives are conflicting: having TALs with few
cells means a larger location update cost but a smaller paging
cost, while having TALs with many cells means a smaller location update cost but a larger paging cost.
Although the TAL scheme can make TA planning more
flexible, it may increase the network complexity and bring
some adverse effects [3]. Since a simple and stable LTE network
is more desirable in the early stage of network construction, we
consider TA planning in a green field, where each TAL has
only one TA in this paper. It is actually a multi-objective optimization problem aiming at finding a rational trade-off
between the paging cost and the location update cost.
Although the two objectives are clear, the details can be very
complicated. TA planning is affected by many other factors,
such as the paging capacity of the mobility management entity
(MME) and geographical features. The first contribution of this
paper is that we build a multi-objective TA planning model by
integrating the network area geographic information. A new
constraint of adjacent cells with no shared boundary crossing
should be assigned to different TAs, is introduced based on the
assumption that each cell must have at least one connected
road to the others. This multi-objective model can provide a set
of trade-off solutions for TA planning, and thus give the decision makers more options, especially taking the anticipated
growth trajectories and technology changes into account.
Evolutionary Multi-objective Optimization (EMO) algorithms are a type of population-based heuristic algorithms,
which use a set of individuals (called population) to search the
Pareto optimal solutions of a multi-objective optimization problem. The main advantage of EMO algorithms over classical
approaches in solving multi-objective optimization problems is
that many trade-off solutions can be obtained in a single run.
Recently, decomposition based EMO algorithms, such as
MOEA/D [28] were reported to achieve good performance in
various application domains [29]. Article with Liu et al. [30]
proposed a new version of MOEA/D by decomposing a multiobjective optimization problem to a number of multi-objective
subproblems (M2M). By M2M decomposition, the population
is decomposed into a number of subpopulations and similar
resources are assigned to optimize each multi-objective subproblem. The second contribution of this paper is the design of a
new M2M-based EMO algorithm for the proposed TA planning model, which lies in two aspects. Firstly, a novel coding
scheme based on the famous four-color theorem [33] is
designed to encode the solutions. A two-step decoding method
based on the coding scheme is designed to decode the solutions.
The first step of decoding tends to merge several small TAs into
a big one while the second step tends to split a big TA into several small ones. The new coding scheme can be beneficial to
balance the two objectives and help to find better trade-off solutions for the multi-objective TA planning model. Secondly, an
M2M decomposition [30] based constraint handling strategy is

Table of Contents for the Digital Edition of Computational Intelligence - February 2017