Royal Road Encodings and Schema Propagation in Selective Crossover

Recombination operators with high positional bias are less disruptive against adjacent genes. Therefore, it is ideal for the encoding to position epistatic genes adjacent to each other and aid GA search through genetic linkage. To produce an encoding that facilitates genetic linkage is problematic. This study focuses on selective crossover, which is an adaptive recombination operator. We propose three alternative encodings for the Royal Road problem. We use these encodings to analyse the performance of selective crossover with respect to different encodings. This study shows that the performance of selective crossover is consistent and is not affected by alternative encodings of a problem, unlike two-point crossover. The encodings are also used to understand the behaviour of selective crossover in terms of schema propagation. Experimental results indicate that selective crossover provides a better balance between exploration and exploitation than conventional recombination operators.