ISSN: 1693-6930
TELKOMNIKA Vol. 14, No. 1, March 2016 : 238 – 244
240
Figure 1. Optimization process
3. Genetic Algorithm
The main characteristic of genetic algorithm is that the strategy of population search and the information exchange between the individuals in the population are not based on
gradient information. It can be used to handle the complicated and non-linear problems which are difficult to be solved by traditional search methods in particular and it can also be widely
applied in such fields as combinatorial optimization, machine learning, self-adaptive control, planning and design as well as artificial life. As a global optimization search algorithm, genetic
algorithm is one of the core intelligent computation technologies in the 21st century for it is easy and universal to apply, it has strong robustness, it can be used in parallel processing and it has
a wide application scope [8].
3.1. The Principles and Methods of Genetic Algorithm
1 Chromosome Encoding Encoding refers to the transformation method to transfer the feasible solutions to one
problem to the search space which genetic algorithm can handle. De Jong once proposed two practical coding principles: one is to use the encoding plan
which is related to the problem to be solved and which has lower-order, short-defined length pattern and the other is to utilize the encoding plan which can give natural presentation or
description to the problem or which has the minimum coded character set [9]. Encoding methods include the following: binary encoding method, gray-code encoding
method, floating-point number encoding method, parameter cascade encoding method and multi-parameter cross-over encoding method.
2 Fitness Computation Basically, there are three methods to transform the objective function value
f x of a
certain point in the solution space to the fitness function value Fit f x
of the corresponding individual in the search space:
a Directly transform the objective function value f x
to be solved to the fitness function value
Fit f x and make
max .
min f x
The objective function is imized
Fit f x f x The objective function is
imize
b As for the minimization problem, perform the following transformation
max max
C f x
f x C
Fit f x others
. Here,
max
C is the maximum input value of
f x .
c If the objective function is the minimization problem, 1
, 0,
1 Fit f x
c c
f x c
f x
2 If the objective function is the maximization problem,
1 ,
0, 1
Fit f x c
c f x
c f x
3
TELKOMNIKA ISSN: 1693-6930
Application of A Self-adaption Dual Population Genetic Algorithm in… Cheng Zhang 241
3.2. The Process of Genetic Algorithm
The genetic operations of genetic algorithm in the entire evolution process are random, but the characteristic it presents is full search. It can effectively use the previous information to
predict the optimization point set with improved expected performance in the next generation. After the continuous evolution from generation to generation, it is finally converged to the
individual which can adapt to the environment at most and the optimal solution to the problem can be obtained. Genetic algorithm involves five elements: parameter coding, setting of initial
population, design of fitness function, design of genetic operation and setting of control parameters [10]. The operations of genetic algorithm are as follows:
1 Selection Selection operation combines elite selection and roulette wheel selection. At first, it
directly copies several elite individuals to the population in the next generation and select the rest individuals with roulette wheel method. In this way, it can not only preserve the excellent
individuals in the population, but also protect the diversity of the individuals in the population. 2 Crossover and Mutation
In the genetic operations, perform crossover and mutation operations at the crossover probability
c
P and mutation probability
m
P . After crossover and mutation operations, conduct
validation test on the newly-generated individuals to check whether the solutions of the new individuals meet the sequence constraints. If so, it proves that these new individuals are
effective; if not, they are invalid and adjustments needs to be made on them. Redistribute some operations to make them effective genes [11, 12].
The basic flowchart of genetic algorithm is indicated as Figure 2.
Figure 2. Basic flowchart of genetic algorithm
4. Design of Multi-objective Genetic Algorithm Based on Self-adaption and Dual Population Strategy