Sub-population policies for a parallel multiobjective genetic algorithm with applications to wing design

In this work a parallel multi-objective genetic algorithm is presented. The population selection and mating phase is kept distinct from the population fitness evaluation loop, that is implemented in parallel. The population can be logically split in subpopulations which number does not depend on the number of processors available for computation. Different sub-population topologies and migration rates among sub-populations can be used. Applications are illustrated both for single and multiobjective mathematical test cases, and aerodynamic transonic wing design.