Hybridizing C-GRASP metaheuristics using the adaptive pattern search method to solve global continuous optimization problems

Abstract

Recently, the interest in solving global continuous optimization problems using metaheuristics has grown in literature. Many of these metaheuristics were originally proposed for combinatorial problems and have been adapted to deal with continuous optimization problems. The Greedy Randomized Adaptive Search Procedure (GRASP) metaheuristic can be included in this group of metaheuristics. Recently, Hirsch et al. (2007) developed the first adaptation of GRASP metaheuristics for the continuous domain, called Continuous-Grasp (C-GRASP). Later, Hirsch et al. (2008) developed a new article and some improvements in C-GRASP were proposed. This work introduces the EC-GRASP (Enhanced Continuous-GRASP) method, a hybrid version composed of CGRASP and the Adaptive Pattern Search (APS) method, a direct search method. The EC-GRASP is a simple method to implement and does not make use of derivate information. The efficiency and robustness of the method are evaluated by comparing EC-GRASP with C-GRASP proposed by Hirsch et al. (2007 e 2008). A set of test functions with global minimum known is used for this comparison. The computational results show the efficiency and robustness of the method. Keywords: C-GRASP; Adaptive Pattern Search; Global continuous optimization.