Modified parallel particle swarm optimization for global optimization using Message Passing Interface

PSO has emerged as a powerful heuristic technique for determining the global optimal solution of nonlinear optimization problems. Like all other evolutionary algorithms (EAs) it is also population based method. However, due to the inherent nature of PSO, it is desirable to parallelize it so as to get a better performance. In this paper, three versions of parallel PSO are presented. They are encoded using the Message Passing Interface (MPI) and are used to solve 16 benchmark scalable test problems available in literature. From the numerical and graphical analysis it is concluded that parallelization helps in enhancing the performance of basic PSO.