Capacity Optimization for UMTS Networks

Abstract
The continuously growing of cellular networks complexity, which followed the introduction of
UMTS technology, has reduced the usefulness of traditional design tools, making them quit
unworthy. The purpose of this project is to illustrate a design tool for UMTS optimized net
planning based on genetic algorithms. In particular, some utilities for 3G net designers, useful
to respect important aspects (such as the environmental one) of the cellular network, are shown.

Keywords: UMTS Network Planning, Evolutionary Algorithms

Introduction to UMTS Networks

The 3rd generation mobile communication system (3G) is put on agenda when the 2nd
generation (2G) digital mobile communication market was significantly evolving. The 2G
mobile communication systems have the following disadvantages: limited frequency spectrum
resources, low frequency spectrum utilization and weak support for mobile multimedia services
(providing only speech and low-speed data services). Also, there was incompatibility between
2G systems. The 2G mobile communication system has a low system capacity hardly meeting
the demand for high speed bandwidth services and impossible for the system to implement
global roaming. Therefore, the 3G communication technology is a natural result in the
advancement of 2G mobile communication technology.
As the internet data services are becoming increasingly popular nowadays, the 3G
communication technology opens the door to a brand new mobile communication world. In
addition to clear voice services, it allows users to conduct multimedia communications with
their personal mobile terminals, for example, internet browsing, multimedia database access,
real time stocks quotes query, videophone, mobile e-commerce, interactive games, wireless
personal audio player, video transmission, knowledge acquisition and entertainments. Some
unique features include location related services, which allows the users to know about their
surroundings anytime, anywhere, for example, block map, locations of hotels, supermarkets
and weather forecasting. The 3G mobile phone has become a good assistant to peoples life
and work.

History
Discussion of a potential successor system for GSM started in ETSI and other standard
developing organizations already in the late 1980, even before any second-generation system
was in commercial operation. The ETSI-term for the future system was Universal Mobile
Telecommunications System (UMTS). Simultaneously, the International Telecommunication
(ITU) also started discussions on a potential future mobile system initially referred to as Future
Public Land Mobile System (FPLMTS) and started to specify a set of system requirements.
Due to the huge world-wide success of GSM, the interest among European network operators
and manufacturers to consider a completely new system was rather low until to the mid-1990s.
Only after the ITU has taken the initiative to formulate a concrete roadmap towards a new

mobile system to be deployed in the early 2000s, the specification activities for UMTS in ETSI
were ramped up in 1995. The ITU term for the future 3G system was later changed to IMT2000, International Telecommunications System for the 2000s. As part of the roadmap, a
deadline for submission of proposals for IMT-2000 by the regional standardization
development organizations was agreed to be in July 1998.In January 1998 ETSI selected two
radio transmission technologies (from originally 4 different proposals) for UMTS terrestrial
radio access (UTRA), referred to as UTRA FDD and UTRA TDD, which were submitted to
ITU as candidates for IMT-2000.
The proposals included a number of different Wideband CDMA (WCDMA) based
Radio access technologies, from ETSI, TTC/ARIB (Japan), TTA (Korea), ANSI T1 (USA) and
TIA (USA), which can be grouped into two types. The one type of proposals requires
synchronized base stations and is building up on the IS-95 2G radio transmission technology.
The other group of concepts does not rely on base station synchronization.

Need of Capacity Optimization

The extraordinary growth of mobile telecommunication sector of the last years has implied
strong economical investments of enterprises that operate in this vital sector, in particular way
from the net infrastructure point of view.
The development of third generation mobile communication (3G) such as UMTS, with the
related advanced allowed services, has increased the need of an efficient network planning
that could keep into account all the aspects of complexity which are typical of this
new technology, changing the traditional approach to this kind of problem.
In fact, even if the WCDMA techniques used by UMTS reduce the problems related to
the frequency management, the capacity of the net represents a vital problem since the
capacity of each radio cell is strongly related to the signal interferenc ratio (SIR), that is a
function of the number and of the kind of active users inside each communication cell .
The need of reduction of radiated power, due to environmental restrictions, and the need of
guaranteeing a good quality of services, requires a capillary distribution of Radio Base
Stations (BSs) on the territory to be covered.
Nowadays, due to the reduced availability of BSs placement zones, it is necessary to seek
new and efficient methods to optimize the cellular coverage services. Different and Interesting
solutions have already been proposed.
One of the most interesting is based on a technique inspired to the natural evolution,
represented by the Genetic Algorithms (GAs), which are good candidates, thanks to their
versatility, to solve a complexand multi-parametric problem such as the considered one.
The purpose of this work is to illustrate a new GAs based method to solve the optimization
coverage and capacity problem of UMTS system, keeping into account its specific features and
typical restrictions found in real situations, such as the environmental one.

Introduction to Evolutionary Algorithms

Search Methods
Search is very comman problem which is included nearly all types of problems. It is as
difficult for computers as for humans. In general, domain space is very huge, and it is
impossible to make a search by trying all possible configurations. There are many search
methods which use different heuristics.

Evolutionary Algorithms
Evolutionary computation uses the computational model of evolutionary processes as key
elements in the design and implementation of computer-based systems and problem solving
applications. There are a variety of evolutionary computational models that have been
proposed and studied which we will refer to as evolutionary algorithms. They share a
common conceptual base of simulating the evolution of individual structures via processes of
selection and reproduction. They depend on the performance (fitness) of the individual
structures. More precisely, evolutionary algorithms maintain a population of structures that
evolve according to rules of selection and other operators, such as recombination and
mutation.
Each individual in the population receives a measure of its fitness in the environment.
Selection focuses attention on high fitness individuals, thus exploiting the available fitness
information. Recombination and mutation perturb those individuals, providing general
heuristics for exploration.
Although simplistic from a biologist's viewpoint, these algorithms are sufficiently complex to
provide robust and powerful adaptive search mechanisms.
A population of individual structures is initialized and then evolved from generation to
generation by repeated applications of evaluation, selection, recombination, and mutation.
The population size N is generally constant in an evolutionary algorithm.
An evolutionary algorithm typically initializes its population randomly, although domain
specific knowledge can also be used to bias the search.
Evaluation measures the fitness of each individual according to its worth in some
environment.
Evaluation may be as simple as computing a fitness function or as complex as running an
elaborate simulation.
Selection is often performed in two steps, parent selection and survival.
Parent selection decides who becomes parents and how many children the parents have.
Children are created via recombination, which exchanges information between parents, and
mutation, which further perturbs the children.
The children are then evaluated. Finally, the survival step decides who survives in the
population.
The origins of evolutionary algorithms can be traced to at least the 1950's.
three methodologies that have emerged in the last few decades:
"evolutionary programming" (Fogel et al., 1966)
"evolution strategies" (Rechenberg, 1973)
"genetic algorithms and genetic programming (Holland, 1975).
Although similar at the highest level, each of these varieties implements an evolutionary
algorithm in a different manner.

The differences include almost all aspects of evolutionary algorithms, including the choices
of representation for the individual structures, types of selection mechanism used, forms of
genetic operators, and measures of performance.

Characteristics of Evolutionary Algorithms

Representation

Selection

Recombination

Mutation

How to define an individual

The way to store the optimization parameters.
Determined according to the problem.
Different types:
Binary representation
Real-valued representation
Lisp-S expression representation

Used to determine parents used to generation of next population

Some types:
Truncation selection
Roulette wheel selection
Tournament selection
Neighborhood selection

Determines how to combine the genes of selected parents

Types is determined according to the representation. These types:
Bits of the genes
Values of the genes

Change on a single gene of the individual

Types is determined according to the representation. These types:
Switching bits
Updating the value of the gene

Fitness Function

Survivor Decision

Gives an intuition about how good the individual is.

Depends on directly to the problem
Idea of survival of the best individuals. It is about Elitism factor.
Usage of it is not advised. In general, overall best individual is stored
as a different individual not to loose reached best parameter set, but
they are not used during other operations.

Evolutionary programming (EP)

Evolutionary programming (EP) developed by Fogel (1966), and traditionally has used
representations that are tailored to the problem domain.
For example, in real-valued optimization problems, the individuals within the population are
real-valued vectors.
Other representations such as ordered lists, and graphical representations could be applied
depending on the problem itself.

procedure EP
{
t = 0;
initialize population P(t);
evaluate P(t);
until (done) {
t = t + 1;
parent_selection P(t);
mutate P(t);
evaluate P(t);
survive P(t);
}
}
After initialization, all N individuals are selected to be parents, and then are mutated,
producing N children.
These children are evaluated and N survivors are chosen from the 2N individuals, using a
probabilistic function based on fitness.
In other words, individuals with a greater fitness have a higher chance of survival.
The form of mutation is based on the representation used.
For example, when using a real-valued vector, each variable within an individual may have
an adaptive mutation rate that is normally distributed with a zero expectation.
Recombination is not generally performed since the forms of mutation used are quite flexible
and can produce perturbations similar to recombination, if desired.

Evolution Strategies (ES)

Evolution Strategies (ES) were independently developed by Rechenberg, with selection,
mutation, and a population of size one.
Schwefel introduced recombination and populations with more than one individual, and
provided a nice comparison of ESs with more traditional optimization techniques.
Evolution strategies typically use real-valued vector representations.
procedure ES; {
t = 0;
initialize population P(t);
evaluate P(t);
until (done) {
t = t + 1;
parent_selection P(t);
recombine P(t)
mutate P(t);
evaluate P(t);
survive P(t);
}
}
After initialization and evaluation, individuals are selected uniformly Randomly to be
parents.
In the standard recombinative ES, pairs of parents produces children via recombination,
which are further perturbed via mutation.
The number of children created is greater than N.
Survival is deterministic and is implemented in one of two ways:

The first allows the N best children to survive, and replaces the parents with these children.
The second allows the N best children and parents to survive.
Like EP, considerable effort has focused on adapting mutation as the algorithm runs by
allowing each variable within an individual to have an adaptive mutation rate that is normally
distributed with a zero expectation. Unlike EP, however, recombination does play an
important role in evolution strategies, especially in adapting mutation.

Genetic Algorithms
Genetic algorithms are considered wide range numerical optimisation methods which use the
natural processes of evolution and genetic recombination. Thanks to their versatility, they
can be used in different application fields.
The algorithms encode each parameters of the problem to be optimised into a proper sequence
(where the alphabet used is generally binary) called a gene, and combine the different
genes to constitute a chromosome. A proper set of chromosomes, called population, undergoes the Darwinian processes of natural selection, mating and mutation, creating new
generations, until it reaches the final optimal solution under the selective pressure of
the desired fitness function.
GA optimisers, therefore, operate according to the following nine points:
1) encoding the solution parameters as genes;
2) creation of chromosomes as strings of genes;
3) initialisation of a starting population;
4) evaluation and assignment of fitness values to the individuals of the population;
5) reproduction by means of fitness weighted selection of individuals belonging to the
populations;
6) recombination to produce recombined members;
7) mutation on the recombined members to produce the members of the next generation.
8) evaluation and assignment of fitness values to the individuals of the next generation;
9) convergence check
The coding is a mapping from the parameter space to the chromosome space and it
transforms the set of parameters, which is generally composed by real numbers, in a string
characterized by a finite length. The parameters are coded into genes of the chromosomes
that allow the GA to evolve independently of the parameters themselves and therefore of the
solution space.
Once created the chromosomes it is necessary choose the number of them which composes the
initial population. This number strongly influences the efficiency of the algorithm in finding
the optimal solution: a high number provides a better sampling of the solution space but
slows the convergence.
Fitness function, or cost function, or object function provides a measure of the goodness of a
given chromosome and therefore the goodness of an individual within a population.
Since the fitness function acts on the parameters themselves, it is necessary to decode
the genes composing a given chromosome to calculate the fitness function of a certain
individual of the population.

Genetic Programming (GP)

It is used to find the proper program for given problem. It requires very high computation
power, so it is suitable for only simple problems. There is no comman representation. But the

most famous representation is the Lisp expression. While the main operator is the cross-over,
mutation is used as secondary operator.
Cross-over method is to replace a gene of the individual with that of another individual.
Individual pairs are selected randomly. Mutation can be applied to only one of the genes or to
whole individual. Meta-Genetic Programming is like to find the program which finds the
program. Even the GP is suitable for simple problems, Meta-Genetic Programming is suitable
for simpler problems.
Because genetic programming and meta-genetic programming are suitable relatively simpler
problems, there are not enough samples for this type of algorithm. Especially meta-genetic
algorithm is a new research area, so documentation on this method is very sparse.
A good sample for genetic programming is:
http://alphard.ethz.ch/gerber/approx/default.html

Definition of the Problem

It is evident that, thanks to their versatility, GAs represent good candidates to solve the typical
optimization problem of UMTS cellular net planning.
GAs have already been used for this kind of problem, even if their application is limited
only to territory coverage. On the contrary, in this paper, other parameters (such as SIR),
that strongly influence the results in real situations, are considered, generating a powerful
tool for optimal net planning.
Some general criteria have been adopted, without reducing the generality of the problem,
which are:
1) it has been considered a suburban area whose dimensions are 3 km x 3 km with
an in homogeneous traffic distribution , even if the proposed algorithm issuitable for
different shaped areas;
2) high gain BSs, placed at the same height, are con-sidered;
3) circular irradiation diagrams of BSs, instead of three lobes diagrams, are considered.
This assumption, made to simplify the implementation of the algorithm, does not
influent the final result;
4) a consolidated electromagnetic propagation model has been adopted;
5) the SIR has been calculated using the following formula,
SIR= SF

Pr
Iin Iout

Figure 1. Operation scheme of a GA iteration.

where SF is the Spreading Factor, Pr is the received power, Iin is the intra-cells interference,
Iout is the inter-cells interference, is the thermal noise.

Proposed Algorithms for Optimization Problem

Since a plenty of goals and restrictions must be respected in a UMTS net, the design can be
made following different criteria. The designer can therefore have different optimization
tools that allow him to consider, in each real situation, the predominant aspects.
For this reason, in this paper, the different mentioned real situations have been considered,
showing the great flexibility of the proposed method.
Case 1
A situation without information about traffic level, without restrictions about the
maximum number of BSs that can be used and without restrictions about their territorial
placement is considered. The goal of this case is the optimization of territorial coverage,
neglecting the performance of service aspects. To reach this target it is necessary to find a
proper fitness function of GA and a proper chromosome.
The BSs are coded, inside the chromosome, by meansof 2 double vectors, that represents
the coordinates of each BS on the territory. To determine the length of the
chromosome, related to the number of considered BSs, the minimum number of BSs
necessary to ensure the coverage of a given percentage pT of the territory, is calculated as:
n_bsmin pT ATot CBS
(2)
where ATot represents the area of the considered territory; pT is the percentage of territory that
must be covered; CBS is the maximum coverage area of each BSs.
Due to the usual not regular shape of the territory to becovered and to the impossibility of per
fectly matching the coverage diagram of near BSs, the value calculated by means of (2) may
be not sufficient and it is necessary to consider a proper multiple n, generally equal to two
In the considered situation, we have n_bsmin=23.
Each gene of the chromosome, representing a BSs, is composed by a number k of variables
equal to 3: 2 are used for the position of the BSs on the territory and 1 is used to represent
the state of activation /deactivation ofthe BSs.
The length of the chromosome ( in term of number of variables) in the considered situation is
expressed by the following formula:
n n_bsmin k

(3)

Case 2
In real situations, the traffic inside a territory is not distributed in a homogeneous way. The
concentration users zones are named hotspots. It is evident that, to guarantee a certain QoS
level, it is necessary to reduce, as more as possible, the intra-cells and inter-cells interference.
As a consequence, placing a BS in a hot spot represents a first significant step in net
optimization.
Given a non homogeneous traffic distribution and aninitial numbers of BSs,
calculated according to (2), the algorithm is capable of maximizing coverage and capacity and
of minimizing cost.
Case 3
In real situation, for environmental reasons, it is not possible to place BSs anywhere.

In this case, only a limited number of zones is available and it is necessary to find a function
that accepts, as inputs, not only information concerning traffic but also information optimize
concerning the available installation zones (in particular their coordinates). The function must
the net considering these limitations that is a cost vinculum. Its structure is therefore equal to
the one of less the cost factor.
Case 4
Another crucial factor in UMTS system is represented by the radiated power (environmental
restrictions), with particular respect to the QoS. Therefore the net needs, sometimes, to
place the BSs on the territory to reduce, as more as possible, the emitted power, guaranteeing an
acceptable level of QoS.
In this case the power of each BS is considered as input parameter (which can be properly
changed), that influences not only the coverage area but also the transmission capacity.

Future Scope
A genetic algorithm based technique to optimize the design of UMTS cellular nets has been
presented.
The proposed method considers most of the limits imposed by the installation of the BSs
necessary to guarantee an optimal service, also including environmental restrictions.
Even if some simplifications were made, the considered technique is capable of ensuring good
results from any point of view, representing a useful tool for UMTS initial optimization.

References
[1]

J. C. S. Cheung, M. A. Beach and J. McGeehan, Network Planning for Third-generation

Mobile Radio Systems, IEEE Communications Magazine, Vol. 32, No. 11,
Nvember 1994, pp. 54-59

[2]

E. Berruto, M. Gudmundson, R. Menolascino, W.Mohr, and M. Pizarroso, Research

Activities on UMTS Radio Interface, Network Architectures, and Planning, IEEE
Communications Magazine, Vol. 36, No. 2, February 1998, pp. 82-95.

[3]

E. Amaldi, A. Capone and F. Malucelli, Planning UMTS Base Station Location:

Optimization Models With Power Control and Algorithms, IEEE Transactions on
Wireless Communications, Vol. 2, No. 5, September 2003, pp. 939-952.

[4]

D. E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning,

Addison Wesley, Massachusetts, 1989.

[5]

L. Nagi and L. Farkas, Indoor Base Station Location Optimization Using Genetic
Algorithms, IEEE International Symposium on Personal, Indoor and Mobile
Communications, Vol. 2, London, 2000, pp. 843-846.

[6]

B. Di Chiara, R. Sorrentino, M. Strappini and L. Tarricone, Genetic Optimization of

Radio Base-station Size and Location Using a GIS-based Frame Work: Experimental
Validation, IEEE International Symposium of Antennas and Propagation Society,
Vol. 2, 2003, pp. 863-866.

[7]

J. K. Han, B. S. Park, Y. S. Choi, and H. K. Park, Genetic Approach with a

New Representation for Base Station Placement in Mobile Communications, IEEE
Vehicular Technology Conference, Vol. 4, No. 54ND, 2001, pp. 2703-2707.

[8]

R. Danesfahani, F. Razzazi and M. R. Shahbazi, An Active Contour Based

Algorithm for Cell Planning, IEEE Internationals Conference on Communications
Technology and Applications, Beijing, 2009, pp. 122-126.