A coevolutionary genetic algorithm for constrained optimization

A co-evolutionary genetic algorithm is proposed for solving constrained optimization problems written as a min-max problem after the introduction of an augmented Lagrangian functional. Two populations are evolved, using in each one, an independent GA. The GA running in population A(B) is a minimization (maximization) one and the individuals in this population encode values of the variable x(y) belonging to the corresponding set X(Y). The GA evolves for a certain number of generations on population A while the other population is kept "frozen". Then the process is applied to population B and the cycle is repeated. The fitness computation is based on the Lagrangian and the fitness of each individual in one population depends on all individuals of the other population. The results of some numerical experiments are presented.