Sunteți pe pagina 1din 4

International Journal of Scientific & Engineering Research, Volume 1, Issue 2, November-2010 1

ISSN 2229-5518

Resource Allocation using Genetic


Algorithms
S. Ranichandra, T. K. P. Rajagopal

ABSTRACT - A genetic-based algorithm can be used to allocate the resources in production units through LPP models. LPP
concentrates on the resource allocations in a more meaningful way so as to make maximum production or to gain the maximum profit or
to minimize the cost of production. There are conventional procedures existing, to solve LPPs. But, in this paper an attempt is made to
solve the problem unconventionally - that is, with the help of Genetic Algorithms (GAs). Here, the drawbacks of the conventional
systems are removed. The benefits of the new system are blindness to auxiliary information, unique algorithm and simple
computerization. The new algorithm is tested with a set of sample problems and the results are found to be satisfactory.

INDEX TERMS – LPP, GA, Crossover, Mutation, Reproduction.

1. INTRODUCTION and finally let


Xj>=0;j = l,2,3,...n … (c)
Many business and economic situations
are concerned with a problem of planning The problem of determining an n-tuple (x1,
activity. In each case, there are limited x2, . . . xn) which makes z a minimum (or
resources at the disposal and the problem is maximum) and which satisfies (b) and (c) is
to make use of these resources so as to yield called the general LPP.
the maximum production or to minimize the
The objective of the GAs is to find an
cost of production, or to give the maximum
optimal solution to a problem. Of course, since
profit, etc. Such problems are referred to as
GAs is heuristic procedures, they are able to
the problems of constrained optimization.
find very good solutions for a wide range of
LPP is a technique for determining an
problems. GAs work by evolving a
optimum schedule of interdependent activity
population of individuals over a number of
in view of the available resources. The term
generations. A fitness value is assigned to each
"programming" means "planning" which refers
individual in the population, where the fitness
to the process of determining a particular
computation depends on the application. The
plan of action from amongst several
main advantages of Genetic Algorithms are:
alternatives. The general form of a LPP is
described as follows. Let z be a linear They are adaptive
function on Rn defined by They have intrinsic parallelism
z = c1x1 + C2X2 + .. . + cnxn … (a) They are efficient for complex
problems
where cj’s are constants. Let (aij) be an m x n
real matrix and let {b1,b2.. . bm} be a set of This paper provides a GA based solution to
constants such that LPP. GAs have been applied to many diverse
a11X1 + a12X2 +.... + a1nXn>= or<=or = or> or<b1 areas such as Function optimization, VLSI
circuit design, Job Shop scheduling, Bin
a21X1 + a22X2 + ... + a2nXn >= or<=or = or> or<b2 packing, Network design, Transportation
problems and etc.
am1x1 + am2X2 + . . . + amnxn >= or<=or = or> or<bm
IJSER © 2010
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 1, Issue 2, November-2010 2
ISSN 2229-5518

2. GENETIC ALGORITHMS (GAS) parent symbol strings forming a new


chromosome string that inherits solution
Genetic Algorithm was invented by characteristics from both parents. Mutation
Professor John Holland at the University of covers this need by injecting new information
Michigan in 1975. It was then made widely in the produced string. The injection is done
popular by Professor David Goldberg at the by randomly altering symbols of the new
University of Illinois. The original Genetic chromosome. Generally mutation is
Algorithm and its many variants are considered as a secondary but not useless
collectively known as Genetic Algorithms. operator that gives a non-zero probability to
GAs are general-purpose search techniques every solution to be considered and
based on principles inspired from the genetic evaluated.
and evolution mechanisms observed in When M new solution strings are
natural systems and populations of living produced, they are considered as a new
beings. Their basic principle is the generation and they totally replace the
maintenance of a population of solutions to a "parents" in order for the evolution to
problem {genotypes) in the form of encoded proceed. Many generations are needed for
information that evolve in time. The the population to converge to the optimum
evolution is based on the laws of natural or a near-optimum solution, the number
selection (survival of the fittest) and genetic increasing according to the problem
information recombination within the difficulty.
population. The evolving population samples
the search space and accumulates knowledge 2.1. SINGLE POINT CROSSOVER
about the good and bad quality areas and Crossover takes two individuals, and
recombining this knowledge to form cuts their chromosome strings at some
solutions with optimal performance to the randomly chosen position, to produce two
specific problem[l]. "head" segments and two "tail" segments. The
At first a population of M solutions tail segments are then swapped over to
does a generated at random, encoded in produce two new full-length chromosomes.
string of symbols (preferably binary strings) Consider A and B as two parents with the
resemble natural chromosomes. Each crossover point as 3.
member of the population is then decoded to
a real problem solution and a "fitness" value PARENT A PARENT B
is assigned to it by a quality function that
10110100 10011110
gives a measure of the solution quality. When
A A
the evaluation is completed, individuals from
the population are selected in pairs to Now headl =101, head2=l 0 0,
replicate and form "offspring" individuals taill=l 0 1 0 0 & tail2=l 1110.
(i.e., new problem solutions). The selection is
performed probabilistically. When two OFFSPRING A OFFSPRING B
parents are selected their symbol strings are
combined to reproduce an "offspring" 10111110 10010 100
solution using genetic-like operators. The A A
main operators used are Crossover and 2.2. TWO POINT CROSSOVER
Mutation. Crossover simply combines the Single point crossover operator has
IJSER © 2010
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 1, Issue 2, November-2010 3
ISSN 2229-5518

some bias of exchange for the right-most bits. 01101011


They have a higher probability of getting 00111101
exchanged than the left-most bits. Take two 00000110
individuals, and cut their chromosome strings 01111100
at two some randomly chosen positions. Notice that all four strings have a 0 in
Interchange the middle portions. This is the left-most bit position of the strings. If the
known as two point crossover. Now consider true optimum solution requires 1 in that
two parents A and B with the crossover points position, then neither reproduction nor
as 3 and 6. crossover can create 1 in that position. Only
mutation can be used to resolve this problem.
PARENT A PARENT B
After mutation is applied on second and
10110100 10011110 fourth strings at first position, we get the
A A A A population as on below.
STRING [After Mutation]
OFFSPRING A OFFSPRING B
01101011
10111100 10010110 00111101
A A A A 00000110
01111100
2.3. REPRODUCTION
In reproduction, good strings in a 3. CONVENTIONAL PROCEDURES VS
population are copied into a large number GENETIC ALGORITHMS
and a mating pool is formed. It is important to
note that no new strings are formed in the LIMITATIONS OF CONVENTIONAL
reproduction phase. PROCEDURES
1. Different Algorithms for
BEFORE AFTER different Methods
REPRODUCTION REPRODUCTION 2. Existence of Conditions and Auxiliary
String _____ Fitness String _____ Fitness Information
10100101 0.0 10111101 0.5 3. Difficulties in Understanding
00110101 0.0 00110101 0.0 the Procedure
10111101 0.5 10111101 0.5 4. Complexities existing in
10011101 0.0 10011101 0.0 Computerization [3].
1010 1100 0.0 1011 1101 0.5
Now the population is in a better form. ADVANTAGES OF USING GENETIC
That is, there are more strings with better ALGORITHMS
fitness. This makes an improvement in 1. Unique Algorithm for any problem
Genetic process. with a change in fitness function
2. Blindness to auxiliary information
2.4. MUTATION and condition
Mutation is the last operator 3. Simple computerization
performed on a bit-by-bit basis. The 4. Simple and uniform procedure [4]
mutation operator changes 1 to 0 and vice
versa. Consider the following population. 4. SOLUTION TO A SAMPLE LPP USING GA
STRING [Before Mutation] SAMPLE PROBLEM
IJSER © 2010
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 1, Issue 2, November-2010 4
ISSN 2229-5518

The following LPP is solved using GA and the surplus variables and artificial variables are to
result is obtained. For this problem, the be introduced. But since GAs is blind to
population is assumed as 100. auxiliary information, the introduction of these
variables is not required. The software is
MAXIMIZE Z = 2 x+3 y capable of accepting all combinations of
Subject to the constraints constraints. This approach attempts to solve
x + y>= 1, x - 5y<= 0, any general LPP, in which any number of
5 x - y >= 0, - x + y>=-l , decision variables can be present. It also
x + y <= 6, y<= 3 And x, y > = 0 solves both maximization and minimization
problems. Since GAs search in a solution
The results obtained are x=3, y=3 and Z=15. space, to get better results a very high
The result is compared with conventional population size is recommended. This work
method and found accurate. provides a better scope to do further research
work in implementing other techniques for
5. COMPUTERIZATION resource allocation.
The computer language used to
develop the software is Java. It is a REFERENCES
revolutionary programming language
[1] David E Goldberg, "Genetic Algorithms in
developed by Sun Microsystems. It is a true
Search, Optimization, and Machine Learning",
object oriented language. The important
Addison-Wesley, 1999.
characteristics are given below [2].
[2] Patrick Naughton and Herbert Schildt,
1. Simple The Complete Reference - "Java™ 2, Tata
2. Object Oriented McGraw-Hill, 1999.
3. Robust
4. Architecture Neutral [3] Srinivas M. and Lalit M Patnaik, "Genetic
5. Multithreaded Algorithms: A Survey", Computer, June 1994.

[4] Tang K S, Man K F, Kwong S and He Q,


6. CONCLUSION "Genetic A;gorithms and their Applications",
In conventional style, either Big-M IEEE Signal Processing Magazine, November
method or Two-Phase simplex method would 1996.
be used to solve the above problem, in which
the auxiliary information like slack variables,

IJSER © 2010
http://www.ijser.org

S-ar putea să vă placă și