Selfish Gene theory and Memetic Algorithms: A fusion of concepts for robust design of hybrid composites

Abstract

The genetic and cultural evolutionary symbiosis of Memetic Algorithms (MAs) has been materialized into the form of a hybrid global-local approach improving both exploration and exploitation properties of search. Instead of local search as performed in MAs, the selfish gene algorithm (SGA) follows a different learning scheme where the conventional population of individuals is replaced by a virtual population of alleles. In this paper a fusion of concepts proposed by MA and SGA is implemented. The proposed approach is a mixed model applying multiple learning procedures aiming to explore the synergy of different cultural transmission rules into the evolutionary process. The principal aspects of approach are: co-evolution of multiple populations, species conservation, migration rules, self-adaptive multiple crossovers, local search in hybrid crossover with local genetic improvements, controlled mutation, individual age control and features-based alleles statistics analysis. Most of these aspects are associated with some kind of problem knowledge and learning from evolution classified as Lamarckian or Baldwinian. In the proposed approach all individuals generated belong inherently to an enlarged population with age structure. Assuming that the age structured population is the virtual population (VP) continuous statistical parameters of alleles population are updating at each generation. Thus, most promising alleles are selected for genes. Then, generation of new individuals following SG theory is based on a pseudo-crossover scheme with changed mating selection and offspring generation mechanisms influenced by best alleles in age-structured VP. Aiming to discuss the capabilities of the proposed approach to deal with robust design optimization of hybrid composite structures a numerical example is presented.