Solving nonlinearly constrained global optimization problem via an auxiliary function method (Q2389996)

Constrained continuous global minimization problems can be converted to unconstrained ones by adding a penalty term to the objective function. Penalty function methods require usually the determination of parameters, so that constrained and unconstrained problems have global minimizers. The authors develop based on the idea of penalty function, a gradient based method for the constrained global optimization problem by introducing an auxiliary function, which does not require any parameters. Asymptotic convergence of the method is proved. Numerical experiments on standard test problems for constrained global optimization problems demonstrate the performance of the method. The algorithm produces low standard deviation for all test problems with not too high computational cost.