Você está na página 1de 4

Review on Real coded Genetic Algorithms

Used in Multiobjective Optimization



Rahila Patel M.M.Raghuwanshi
R.C.E.R.T., Chandrapur (m.s.),
India
NYSS College of Engineering and
Research, Nagpur (m.s.), India
rahila.patel@gmail.com m_raghuwanshi@rediffmail.com

Abstract: This paper gives a short review of real coded
genetic algorithm (RCGA) used for multiobjective
optimization. Handling of continues search space is very
easy with RCGA and solution representation is very close to
natural formulation of real-world problems. Because of the
obvious reasons, most of real-world multi-objective
optimization problems are solved using RCGA. The topics
discussed in this paper include new algorithms, design issues
of multi-objective optimization like efficiency, scalability,
constraint handling and self-adaptation. This discussion
suggests potential areas for future research, namely, design
of new algorithm, new recombination operator and Pareto
optimal front formation techniques.

Keywords-Real-Coded genetic algorithm (RCGA), Multi-
objective Evolutionary Algorithm (MOEA), Evolutionary
Algorithm (EA), Multi-objective optimization, Evolutionary
Multi-objective optimization (EMO).

1. Introduction
Genetic algorithms are search and optimization tool
inspired by principles of natural genetics. Various single
objective and multi objective problems have been solved
using genetic algorithm. John Holland, University of
Michigan, Ann Arbor, first conceived the concept of
genetic algorithm [1,2]. In Real-coded GA, decision
variables are used directly (without coding) to form
chromosome-like structure. Chromosome represents a
solution and population is a collection of such solutions.
The operators (selection, recombination and mutation)
modify the population of the solution to create new (and
hopefully better) population. [3,4]
Handling of continues search space is very easy with
RCGA and solution representation is very close to natural
formulation of real-world problems. Because of the
obvious reasons, most of real-world multi-objective
optimization problems are solved using RCGA.[4,5,6]
The doctoral study on vector evaluated genetic algorithm
(VEGA) by Dave Schaffer in 1984 [7] and Goldbergs
suggestion for the use of non-dominated sorting along
with a niching mechanism [1] generated an
overwhelming interest on MOEAs.
The two fundamental goals in MOEA design are
guiding the search towards the Pareto set and keeping a
diverse set of non-dominated solutions. It is also
important to achieve these goals in a computationally fast
manner. The first generation of MOEA was characterized
by use of selection mechanism based on Pareto ranking.
Fitness sharing was the most common approach to
maintain diversity of the same. The second generation of
MOEA can be characterized by an emphasis on efficiency
and by the use of elitism. This generation also addresses
the issue of test problems and different metrics to
measure performance of MOEA. Attempts are made to
develop theoretical foundation of MOEA[8]. MOEA
literature contains number of survey articles describing
the state-of-the-art [5]-[14].
The paper is organized as follows. Section 2
summarizes basic principles of multiobjective
optimization. Section 3 gives overview of Real coded
Genetic Algorithm used for Multiobjective Optimization.
Section 4 focuses on algorithm design issues and presents
concepts and techniques that have been developed to deal
with the additional complexity caused by multiple
objectives. Section 5 is on conclusions and challenges as
future tasks.

2. Multiobjective Optimization
The Multiobjective Optimization problem in its
general form can be described as follows:
Minimize/Maximize m=1,2,..,M; ), (x f
m
Subjected to j=1,2,..,J; 0 ) ( x g
j
=0 k=1,2,..,K; ) (x h
k
, i=1,2,...,n.
) ( ) ( u
i i
L
i
x x x
A solution x X that satisfy all the (J+K) constraints
and all of the 2N variable bounds stated above is called
feasible solution. X F S
n
R , where F is a feasible
region in search space S. Objective space Z=f (X) is the
image of the decision space X under the objective
function f. The superiority (or dominance) of one solution
over the other cannot be established with many objectives
in mind. Objective vectors are compared according to the
dominance relation defined below
Definition 1 (Dominance relation): x, y
n
R then x is
said to dominate y, denoted as y x iff for
i i
y x
i .and
j j
y x j :
Definition 2 (Pareto set): Let F
n
R be a set of
vectors. Then the Pareto set F* F
is defined as F*={x F | y F: y x}.
Vectors in F* are called Pareto vectors of F. For given set
F, the set F* is unique. Moreover, for a given set F, the
set F* is of substantial size.
Third International Conference on Emerging Trends in Engineering and Technology
978-0-7695-4246-1/10 $26.00 2010 IEEE
DOI 10.1109/ICETET.2010.112
610
Third International Conference on Emerging Trends in Engineering and Technology
978-0-7695-4246-1/10 $26.00 2010 IEEE
DOI 10.1109/ICETET.2010.112
610
Third International Conference on Emerging Trends in Engineering and Technology
978-0-7695-4246-1/10 $26.00 2010 IEEE
DOI 10.1109/ICETET.2010.112
610
Third International Conference on Emerging Trends in Engineering and Technology
978-0-7695-4246-1/10 $26.00 2010 IEEE
DOI 10.1109/ICETET.2010.112
610
In the search space non-dominated solutions are
known as Pareto-optimal solutions. The curve joining
these solutions is known as a Pareto-optimal front. There
are two goals in a multiobjective optimization: To find a
set of solutions as close as possible to the Pareto optimal
front and to find a set of solutions as diverse (or non-
dominated) as possible.[4]
3. Real coded Genetic Algorithm used for
Multiobjective Optimization
In Real-coded GA, decision variables are used directly
(without coding) to form chromosome-like structure. Real
coded GA requires a special recombination and mutation
operators. [8] [14]
Here we present a short chronological survey of the real
coded genetic algorithm used for multiobjective
optimization during the last two decades in terms of non-
elitist and elitist algorithms.
3.1 Non-Elitist Multi-Objective Evolutionary Algorithms
Schaffers vector-evaluated genetic algorithm
(VEGA): Schaffer modified the simple three-operator
genetic algorithm (with selection, crossover and
mutation) by performing independent selection cycles
according to each objective. The selection method is
separated for each individual objective to fill up a portion
of the mating pool. Then the entire population is
thoroughly shuffled to apply crossover and mutation
operators. The algorithm worked efficiently for some
generations but in some cases suffered from its bias
towards some individuals or regions.[7]
Multi-Objective Genetic Algorithm (MOGA):This
algorithm is suggested by Fonseca and Fleming in [15].
In MOGA, the rank of a certain individual corresponds to
the number of chromosomes in the current population by
which it is dominated. All non-dominated individuals are
assigned the highest possible fitness value, while
dominated ones are penalized according to the population
density of the corresponding region to which they belong
(i.e., fitness sharing is used to verify how crowded is the
region surrounding each individual).
Niched-Pareto Genetic Algorithm (NPGA): This
algorithm is suggested by Horn, Natpliotis and Goldberg
in [16]. The NPGA uses a tournament selection scheme
based on Pareto dominance. The basic idea of the
algorithm is: two individuals are randomly chosen and
compared against a subset from the entire population
(typically, around 10% of the population). If one of them
is dominated (by the individuals randomly chosen from
the population) and the other is not, then the non-
dominated individual wins.
Non-dominated Sorting Genetic Algorithm
(NSGA):The algorithm is suggested by Srinivas and Deb
[17]. Before selection is performed, the population is
ranked on the basis of non-domination: all non-dominated
individuals are classified into one category. To maintain
the diversity of the population, these classified
individuals are shared with their dummy fitness values.
The algorithm of the NSGA is not very efficient, because
Pareto ranking has to be repeated over and over again.
3. 2. Elitist Multi-Objective Evolutionary Algorithms
The elitist EMO methodologies include an elite-
preservation mechanism in their procedures.
The wide development of EMOO algorithms in the recent
years has begun after the works of Eckart Zitzler [18],
due to it the elitism has become a standard mechanism in
the development in this direction.
Non-dominated Sorting Genetic Algorithm II
(NSGA-II): NSGA-II is introduced by Deb and Agarwal
[19] as an improved version of the NSGA . In NSGA-II.
for each solution one has to determine how many
solutions dominate it and the set of solutions to which it
dominates. The NSGA-II estimates the density of
solutions surrounding a particular solution in the
population by computing the average distance of two
points on either side of this point along each of the
objectives of the problem. This value is the so-called
crowding distance. During selection, the NSGA-II uses a
crowded-comparison operator which takes into
consideration both the non-domination rank of an
individual in the population and its crowding distance.
The elitist mechanism of the NSGA-II consists of
combining the best parents with the best offspring
obtained. Its mechanism is better.
Strength Pareto Evolutionary Algorithm (SPEA):
This algorithm is introduced by Zitzler and Thiele in
[18]. SPEA uses an archive containing non-dominated
solutions previously found. At each generation, non-
dominated individuals are copied to the external
nondominated set. For each individual in this external set,
a strength value is computed.
Strength Pareto Evolutionary Algorithm 2 (SPEA
2):Proposed by zitzler, it has three main differences with
respect to its predecessor SPEA: (1) it incorporates a fine-
grained fitness assignment strategy which takes into
account for each individual the number of individuals that
dominate it and the number of individuals by which it is
dominated; (2) it uses a nearest neighbor density
estimation technique which guides the search more
efficiently; (3) it has an enhanced archive truncation
method that guarantees the preservation of boundary
solutions. [20]
Pareto Archived Evolution Strategy (PAES)
Algorithm: This algorithm is introduced by Knowles and
Corne [21]. PAES consists of a 1 + 1 evolution strategy
(i.e., a single parent that generates a single offspring) in
combination with a historical archive that records the
non-dominated solutions previously found. A special
feature of this algorithm is the procedure used to maintain
diversity, which consists of a crowding procedure that
divides objective space in a recursive manner. Since the
procedure is adaptive, no extra parameters are required.
Pareto Envelope based Selection Algorithm (PESA)
: Corne et al. [22] suggested an algorithm known as
PESA, which combines the good aspects of SPEA and
PAES. Like SPEA, PESA carries two populations (a
smaller EA population and a larger archive population).
Non-domination and the PAES crowding concept is used
to update the archive with the newly created child
solutions.
611 611 611 611
In an extended version of PESA [23], instead of
applying the selection procedure on population members,
hyper boxes in the objective space are selected based on
the number of solutions residing in the hyper boxes. This
region-based selection procedure has shown to perform
better than the individual-based selection procedure of
PESA.
4. Design issues of Multiobjective Genetic
Algorithm
The primary questions when developing genetic
algorithms for multi-objective problems are how to
evaluate fitness, how to determine which potential
solution points should be passed on to the next
generation, and how to incorporate the idea of Pareto
optimality. The approaches that are described in this
subsection collectively address these issues.
Fitness assignment : There are various strategies for
fitness assignment. The aggregation-based strategy,
aggregate the objectives into a single parameterized
objective e.g. weighted-sum aggregation. Criterion based
methods switch between the objectives during the
selection phase. Pareto dominance based methods use
different approaches like dominance rank, dominance
depth and dominance count. The dominance count is the
number of individuals dominated by a certain
individual.[8]
Selection: Selection operator has got the force that
may pull the search to a narrow area of search space or it
may lend the algorithm to search the entire search space.
Most MOEAs try to maintain diversity within the current
Pareto set approximation by incorporating density
information into the selection process. The NPGA uses
binary tournament selection. The NPGA never ranks an
individual with respect to the entire population. This
results in a faster algorithm. Another remarkable work in
the same direction of the NPGA is the improved ranking
procedure proposed by Jensen [24], which significantly
reduces the computational complexity of the NSGA-II.
Objective reduction: Nowadays, a more common
research trend has been to focus on the design of MOEAs
that reduce the number of objective function evaluations
performed, under the assumption that such evaluations
may be very expensive in some real-world applications.
For that sake, MOEA researchers have been adopting
techniques such as surrogate models, which have long
been used in engineering [25]. The main idea of surrogate
models is to build an approximate model of the problem,
which is cheap to evaluate.
Scalability: The deterioration of the performance of
evolutionary algorithms (EAs) with increasing number of
objectives has already been identified and reported in
literature [26]. It is recognized that with an increase in the
number of objectives, there is a substantial reduction in
the number of fronts, with each front becoming denser.
Given this, most solutions are assigned the same rank of
non-dominance and Pareto selection weakens since it has
to discriminate mostly based on the diversity of solutions.
The remedial approaches proposed are collectively
referred as preference ordering based approaches. These
approaches primarily include:
Relaxed form of Pareto-dominance [26]
Controlling the Dominance Area [27]
Modification of rank definition [28]
Utilization of Scalarizing fitness functions [29]
Indicator based EMO algorithm [30]
Substitute distance metrics [31]
Constraint-handling: One of the research areas that
has attracted a lot of interest in recent years has been the
use of multi-objective optimization concepts to design
constraint-handling mechanisms for EAs [32]. Most of
the current work has focused on extending the Pareto
optimality relation in order to incorporate constraints.
Also, the use of penalty functions that punish a solution
for not being feasible are easy to incorporate into a
MOEA.
Incorporation of users preferences: In practical
applications of MOEAs, users are normally not interested
in a large number of nondominated solutions. Instead,
they are usually only interested in a few types of tradeoffs
among the objectives. Thus, if such users preferences are
incorporated into the selection mechanism of a MOEA,
the search can be much more efficient and the results
more meaningful. Although some research has been done
in this direction [33], it is still relatively uncommon to
report results of a MOEA that incorporates users
preferences.
Self- adaptation: The design of mechanisms that
allow an automated control of the parameters of a MOEA
(by using, for example, self-adaptation [34], so that the
MOEA can adapt its parameters without any human
intervention) has been scarcely explored by EMO
researchers.
5. Conclusion
This paper has reviewed multi-objective optimization
using real-coded genetic algorithm. Right from historical
analysis of the development to the design issues,
challenges and new techniques of multi-objective
optimization evolutionary algorithms are discussed here.
We discussed various methods and theory on MOEA.
Basic principles of multiobjective optimization and EA
are presented and various algorithmic concepts such as
fitness assignment, diversity preservation and elitism are
discussed. Nothing much has been found in the literature
about use of different recombination operator.
This survey suggests a number of tasks, which are outline
in the following:
1. Design of new schemes of sub population (niche)
formation for multiobjective optimization real-coded
genetic algorithm
2. Use of more efficient data structures to store non-
dominated vectors.
3. Design and implementation of niching genetic
algorithm for multiobjective optimization problems
5. Study on why when and how multi-parent
recombination works better than two parent
recombination for multiobjective optimization real-coded
genetic algorithm
612 612 612 612

REFERENCES
1. Goldberg D E., Genetic Algorithms in Search, Optimization and
Machine Learning, Reading: Addison-Wesley Publishing Company,
1989
2. Eiben A E, Smith J E., Introduction to Evolutionary Computing,
Berlin: Springer, 2003
3. Coello Coello C A, Lamont G B, Van Veldhuizen D A. 2nd ed.,
Evolutionary Algorithms for Solving Multi-Objective Problems,
New York: Springer, 2007
4. Deb K., Multi-Objective Optimization using Evolutionary
Algorithms, Chichester: John Wiley & Sons, 2001
5. Coello Coello C A., An updated survey of GA-based multiobjective
optimization techniques, ACM Computing Surveys, 2000, 32(2):
109143
6. Miettinen K M., Nonlinear Multiobjective Optimization, Boston:
Kluwer Academic Publishers, 1999
7. Schaffer J D., Multiple objective optimization with vector evaluated
genetic algorithms,PhD thesis. Nashville: Vanderbilt University,
1984
8. M.M.Raghuwanshi, O.G.Kakde, Survey on multiobjective
evolutionary and real coded genetic algorithms, on-line journal of
Complexity International (vol. 11)] 2005
9. Coello Coello C A , Evolutionary multi-objective optimization :
Some current research trends and topics that remains to be explored,
Front. Comput. Sci. China, 3(1): 1830, 2009
10. Dhish Kumar Saxena, Tapabrata Ray, Kalyanmoy Deb and
Ashutosh Tiwari, Constrained Many-objective Optimization: A way
forward, IEEE978-1-4244-2959 -2/09/$25.00 2009
11. H. Ishibuchi, N. Tsukamoto, and Y. Nojima, Evolutionary many-
objective optimization: A short review, Proc. of 2008 IEEE
Congress on Evolutionary Computation, pp. 2424-2431, Hong Kong,
June 1-6, 2008.
12. Vassil Guliashki, Hristo Toshev, Chavdar Korsemov, Survey of
Evolutionary Algorithms Used in Multiobjective Optimization,
PROBLEMS OF ENGINEERING CYBERNETICS AND
ROBOTICS, 60 BULGARIAN ACADEMY OF SCIENCES. 2009
13. Kalyanmoy Deb, Current trends in evolutionary multi-objective
optimization, Int. J. Simul. Multidisci. Des. Optim. 1, 18 , 2007
14. Eckart Zitzler,Lothar Thiele, Marco Laumanns, Carlos M. Fonseca,
and Viviane Grunert da Fonseca, Performance Assessment of
Multiobjective Optimizers: An Analysis and Review, IEEE
TRANSACTIONS ON EVOLUTIONARY COMPUTATION, VOL.
7, NO. 2, APRIL 2003
15. Fonseca, C. M., P. J. Fleming, Genetic Algorithms for
Muitiobjective Optimization: Formulation, Discussion and
Generalization, In: Proc. of the 5th International Conference on
Genetic Algorithms, San Mateo, California, University of Illinois at
Urbana-Champaign (Stephanie Forrest, Ed.), Morgan Kauffman
Publishers, 416-423, 1993.
16. Horn, J., N. Nafpliotis, D.E. Goldberg, A Niched Pareto Genetic
Algorithm for Multiobjective Optimization, In: Proc. of the First
IEEE Conference on Evolutionary Computation, , 82-87, June
1994.
17. Srinivas, N., K. Deb, Multiobjective Optimization Using
Nondominated Sorting in Genetic Algorithms, Evolutionary
Computation, Vol. 2, No 3, 221-248, 1994.
18. Zitzler, E., L. Thiele, Multiobjective Evolutionary Algorithms: A
Comparative Case Study and the Strength Pareto Approach, In:
IEEE Transactions on Evolutionary Computation, Vol. 3,
November 1999, No 4, 257-271, 1999.
19. Deb, K., A. Pratap, S. Agarwal, T. Meyarivan, A Fast and Elitist
Multiobjective Genetic Algorithm: NSGA-II, IEEE Transactions
on Evolutionary Computation, Vol. 6, No 2, 182-197. Apr. 2002
20. Zitzler, E., M. Laumanns, L. Thiele, SPEA2: Improving the
Strength Pareto Evolutionary Algorithm, In: EUROGEN 2001.
21. Knowles, J. D., D. W. Corne, Approximating the Nondominated
Front Using the Pareto Archived Evolution Strategy, Evolutionary
Computation, Vol. 8, No 2, 149-172, 2000.
22. Corne, D. W., J. D. Knowles, M. J. Oates, The Pareto Envelope-
Based Selection Algorithm for Multi-Objective Optimization, In:
PPSN 2000. LNCS, Vol. 1917, Springer, Heidelberg (K. Deb, et al.,
Eds.), 839-848. , 2000
23. Corne, D. W., J. D. Knowles, M. J. Oates, PESA-II: Region-Based
Selection in Evolutionary Multiobjective Optimization. In: Proc.
of the Genetic and Evolutionary Computation Conference
,GECCO2001
24.Jensen M T., Reducing the run-time complexity of multionbjective
EAs: the NSGA-II and other algorithms, IEEE Transactions on
Evolutionary Computation, 7(5): 503515, 2003
25. Knowles J., ParEGO:A hybrid algorithm with on-line landscape
approximation for exersive multiobjective optimization problems,
IEEE Transactions on Evolutionary Computation, 10(1): 5066,
2006
26. K. Ikeda, H. Kita, and S. Kobayashi. Failure of pareto-based moeas:
Does non-dominated really mean near to optimal? In IEEE
Congress on Evolutionary Computation, volume 2, pages 957962,
2001.
27 .Sato, H. E. Aguirre, and K. Tanaka. Controlling dominance area
of solutions and its impact on the performance of moaes. 4th
International Conference, EMO 2007, pages 520, Matshushima,
Japan, March 2007.
28 D. W. Corne and J. D. Knowles. Techniques for highly
multiobjective optimisation: some nondominated points are better
than others. In Genetic and Evolutionary Computation Conference
(GECCO), pages 773780, 2007.
29 H. Ishibuchi, T. Doi, and Y. Nojima., Incorporation of scalarizing
fitness functions into evolutionary multiobjective optimization
algorithms, In Proceedings of the Parallel Problem Solving from
Nature IX (PPSN-IX), pages 493502. Berlin: Springer, 2006.
30. E. Zitzler and S. Kunzli., Indicator-based selection in
multiobjective search, In PPSN VIII, volume 3242, pages 832842.
Springer LNCS, 2004.
31. K. Singh, A. Issac, T. Ray, and W. Smith., A study on the
performance of substitute distance based approaches for
evolutionary many objective optimization, In Simulated Evolution
and Learning). Germany: Springer, 2008.
32. Mezura-Montes E, Coello Coello C A., Constrained optimization
via multiobjective evolutionary algorithms, In: Knowles J, Corne
D, Deb K, eds. Multi-Objective Problem Solving from Nature:
From Concepts to Applications. Berlin: Springer, 5375, 2008
33. Branke J, Deb K., Integrating user preferences into evolutionary
multiobjective optimization, In: Jin Y C, ed. Knowledge
Incorporation in Evolutionary Computation. Berlin: Springer, 461
477, 2005
34. Meyer-Nieberg S, Beyer H G., Self-adaptation in evolutionary
algorithms, In: Lobo F G, Lima C F, Michalewicz Z, eds.
Parameter Setting in Evolutionary Algorithms. Berlin: Springer-
Verlag,4775, 2007
613 613 613 613

Você também pode gostar