The gravitational search algorithm (GSA) is a novel heuristic method inspired by Newton's gravity and velocity equations. In addition, it is a population-based algorithm, in which each member (called an agent) in the population has a mass, velocity, and acceleration. Beginning with the first population state, agents influence each other via mass and velocity relations. This mutual effect causes agents to reach the optimum. Hence, the performance of the GSA to attain the optimum is related to the initial population formation, like other population-based algorithms. In this study, the orthogonal array (OA) concept is applied and injected to the GSA algorithm in the initialization phase. Hence, the GSA benefits from the homogenized agent distribution tendency of the OA. The implementation results are utilized to compare the conventional and proposed methods (i.e. conventional GSA and the so-called "OA-GSA"), and the efficiency of the proposed method is demonstrated.