In this paper, the improvement of Particle Swam Optimization (PSO) is proposed. PSO is algorithm created by mimicking the food-seeking behavior of swarm of organisms, such as birds and fish. In recent years, PSO is drawing much attention as one of the evolutionary computation methods to obtain the approximate optimal solution for the continuous optimization problem with multi-peak objective function. But, one of the major weaknesses of PSO is trapped into local optima. To overcome this weakness, this paper proposes the new strategy of information sharing called Rbest model and the parameter adjustment with consideration to searching phase and state (Application of the mutation concept). Then, in the computational experiment, the benchmark problems are tested in order to validate the effectiveness of proposed method (2n-minima: 61% improvement from the original PSO, Rastrigin: 8% improvement from the AFPSO)  .