Performance Evaluation of ES Type Genetic Algorithms for Solving Block Layout Problems with Floor Constraints (Self-Tuning Method for GA Parameters)

摘要

This paper deals with the problem of 2 D facility layout planning, in which each placement unit is a rectangular block having various shapes. In actual facility planning, a strong constraint might be added to a placement region. We cannot disregard the constraint. We propose some algorithms based on the evolution strategy (ES) and the genetic algorithm (GA) for the layout problem with the constraint. After tuning major parameters of the GA method and a family of the ES type algorithms, we compare the performances of these algorithms. Especially, we clarify effects of a multiple point mutation method and a new mutation method using normal random numbers on the ES type algorithms. Also, we propose a self-tuning method of parameters for an ES type algorithm and show effectiveness of this method based on numerical experiments.