Abstract: By analyzing mechanism of premature phenomenon in particle swarm optimization (PSO), we found nature of premature convergence and proposed a "leap" strategy to jump out of local minimum, making halted particles "renewed" when they are trapped into a local optimum. The strategy is applied to nonlinear programming and results are encouraging. The improved PSO solves efficiently premature convergence of the algorithm applying in nonlinear optimizations and improves global search ability of PSO.

Key words: particle swarm optimization, premature converge, systems engineering, "leap"strategy