Research On Full Dominance-Recessive Diploid Genetic Algorithm In Multi-Objective Structural Optimization

This paper presents a single-locus tetra-allelic binary Genetic Algorithm based on the concept of dominant and recessive genes, which is named Full Dominance-Recessive Diploid GA. This GA features separating dominant one-zero with recessive one-zero. Numerical experiments with four test functions are performed, and some valuable conclusions are reached. First, the efficacy of Full Dominance-Recessive Diploid Genetic Algorithm outperforms the Standard Genetic Algorithm (SGA) in high-dimensional functions. The time complexity of the proposed GA approximates a polynomial multiple of that of SGA. The analysis on convergence and diversity shows that architecture of GA has great impact on convergence, and premature convergence seems to have little relation with lack of diversity. It is also found that the average solution of each iteration should be an important indicator of GA. Comparison and contrast on search space of the SGA and Full Dominance-Recessive Diploid GA justifies the latter's capability to search hi a larger representation space. When applied to practical multi-objective design optimization, the proposed GA obtains a set of desired Pareto solutions. Finally, a summary is given together with suggestions for future research.