Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
98 views

Using The Min-Max Method To Solve Multiobjective Optimization Problems With Genetic Algorithms

This document discusses using genetic algorithms to solve multi-objective optimization problems. It introduces a new approach based on the concept of min-max optimum from operations research. This technique can produce the Pareto set and best trade-offs among objectives. The results from this method are compared to other mathematical programming techniques and genetic algorithm approaches. The importance of representation is also discussed, as reducing the length of chromosomes tended to produce better results, even at the expense of using a higher cardinality alphabet.

Uploaded by

Nguyen Vinh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
98 views

Using The Min-Max Method To Solve Multiobjective Optimization Problems With Genetic Algorithms

This document discusses using genetic algorithms to solve multi-objective optimization problems. It introduces a new approach based on the concept of min-max optimum from operations research. This technique can produce the Pareto set and best trade-offs among objectives. The results from this method are compared to other mathematical programming techniques and genetic algorithm approaches. The importance of representation is also discussed, as reducing the length of chromosomes tended to produce better results, even at the expense of using a higher cardinality alphabet.

Uploaded by

Nguyen Vinh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Using the Min-Max Method to Solve

Multiobjective Optimization Problems with


Genetic Algorithms

Carlos A. Coello Coello

Engineering Design Centre


University of Plymouth
Plymouth, Devon PL4 8AA, UK
ccoello@soc.plym.ac.uk

Abstract. In this paper, a new multiobjective optimization technique


based on the genetic algorithm (GA) is introduced. This method is based
in the concept of min-max optimum, taken from the Operations Research
literature, and can produce the Pareto set and the best trade-off among
the objectives. The results produced by this approach are compared to
those produced with other mathematical programming techniques and
GA-based approaches using a multiobjective optimization tool called
MOSES (Multiobjective Optimization of Systems in the Engineering Sci-
ences). The importance of representation is hinted in the example used,
since it can be seen that reducing the chromosomic length of an individ-
ual tends to produce better results in the optimization process, even if
it’s at the expense of a higher cardinality alphabet.

1 Introduction

Engineering optimization has been a very fertile area of research in the last few
years, but the normal trend has been to deal with a single objective at a time, and
use ideal and unrealistic problems, rather than real-world applications. Assuming
only one objective is generally unrealistic for engineering optimization problems,
since most real-world problems have several (possibly conflicting) objectives. The
common practice, therefore, has been to let the designer to make decisions based
on his/her experience, instead of using some well-defined optimality criterion.
Over the years, the Operations Research community has produced more
than 20 mathematical programming techniques to deal with multiple objectives.
However, the main focus of these approaches is to produce a single trade-off based
on some notion of optimality, rather than producing several possible alternatives
from which the designer may choose. More recently, the genetic algorithm (GA),
an artificial intelligence search technique based on the mechanics of natural se-
lection, has been found to be effective on some scalar optimization problems. In

Part of this work was developed by the author while visiting LANIA (Laboratorio
Nacional de Informática Avanzada) in Xalapa, Veracruz, México

Helder Coelho (Ed.): IBERAMIA’98, LNAI 1484, pp. 303–313, 1998.



c Springer-Verlag Berlin Heidelberg 1998
304 Carlos A. Coello Coello

order to extend the GA to deal with multiple objectives, the structure of the
GA has been modified to handle a vector fitness function.
This paper will review some of the previous work in multiobjective optimiza-
tion using GAs, and a new approach, proposed by the author, will be introduced.
Also, MOSES (Multiobjective Optimization of Systems in the Engineering Sci-
ences), a system developed as a testbed for multiobjective optimization tech-
niques by the author, will be briefly described together with an example of its
use. The new approach, based on the notion of min-max optimum, is able to
generate the Pareto set and better trade-offs than any of the other techniques
included in MOSES. The importance of using alphabets of cardinality higher
than two will be emphasized, and the results found with this alternative rep-
resentation will be shown to be better than those produced using a traditional
binary representation, both for single and multiobjective optimization.

1.1 Statement of the Problem

Multiobjective optimization (also called multicriteria optimization, multiperfor-


mance or vector optimization) can be defined as the problem of finding [13]:

a vector of decision variables which satisfies constraints and optimizes a


vector function whose elements represent the objective functions. These
functions form a mathematical description of performance criteria which
are usually in conflict with each other. Hence, the term “optimize” means
finding such a solution which would give the values of all the objective
functions acceptable to the designer.

Formally, we can state it as follows:

Find the vector x̄∗ = [x∗1 , x∗2 , . . . , x∗n ]T which will satisfy the m inequality
constraints:

gi (x̄) ≥ 0 i = 1, 2, . . . , m (1)
the p equality constraints

hi (x̄) = 0 i = 1, 2, . . . , p (2)
and optimize the vector function
T
f¯(x̄) = [f1 (x̄), f2 (x̄), . . . , fk (x̄)] (3)
T
where x̄ = [x1 , x2 , . . . , xn ] is the vector of decision variables.

1.2 Min-Max Optimum

The idea of stating the min-max optimum and applying it to multiobjective


optimization problems, was taken from game theory, which deals with solving

You might also like