Constructive Design and Evauation of Genetic Algorithm for Scheduling Problem of Robot Control Computation

点击翻译按钮获取中文摘要

We studied practical methodology to design an effective genetic algoritym for the scheduling problem of robot control computation. Then, we proposed the following techniques applicable to the implementation of genetic algorithms for many combinatorial optimization problems.(1) Encoding of isomorphic genotype : A phenotype is usually represented by several genotypes. We have proposed a new technique to represent each phenotype uniquely by using a set of isomorphic genotypes. We could maintain the diversity of population with the encoding technique.(2) Phenotypic Distance : We have defined the phenotypic distance between two phenotypes by the least Hamming distance between isomorphic genotypes. By using the phenotypic distance in genetic operations, we could evaluate and control the diversity of population.(3) Weight-Edge Crossover : We have proposed a new crossover which combines the conventional edge crossover with a heuristic insight to preserve the excellent characteristic of parents. Experimental results showed that the proposed crossover was superior to the conventional ones.