| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系所 === 99 === The multi-objectives optimization problem has a quite different point of view compared with a single objective problem. There is only one global optimum solution in the single objective optimization problem, but there is a set of solutions in multi-objectives optimization problem, called the Pareto-optimal set. This set constitutes global optimum solutions; which are considered to be equally important and non-dominated with each other.
This paper describes a genetic algorithm based multi-objectives optimization problem that uses sequence pair as the data structure of chromosomes and incorporates the concept of archive in order to deal properly with the multi-objectives problem. The difference in genetic algorithm between the traditional one and ours is we forsake crossover operation and focus on mutation operation, combining the idea of orthogonal experiment design with swap mutation operator. For the purpose of achieving global optimal, we add the concepts of simulated annealing algorithm that accepting the inferior solutions at the beginning.
The first step in the procedure for executing traditional genetic algorithm is to generate the fix number of chromosomes for crossover and mutation operation. For the purpose of expanding a diversity of initial population, we generate a large amount of chromosomes; besides, only have a part of chromosomes will be selected as initial population. In order to increase efficiency of genetic algorithm in searching solution space and significantly decrease the mutation operation in un-ideal region, we dynamically adjust mutation times as generation rise and fine-tune each generation’s mutation times after each mutation operation.
|
|---|
