A hybrid genetic algorithm for nonlinear bilevel programming

For nonlinear bilevel programming whose leader is a nonlinear function, a hybrid genetic algorithm together with a simplex algorithm is presented. In order to tackle the problems of canonical genetic algorithm such as premature convergence, lower convergent speed to global optimization and lower precision solutions, the gradient projection method is used to design an effective direction mutation operator such that the operator can produce better offspring. The numerical simulation also shows that the new algorithm is very effective.