The particle swarm optimization (PSO) is a stochastic population-based optimization technique that is gaining popularity. However, it may be trapped in local optima and fail to converge to global optimum, especially for multimodal and high-dimensional problems. Many variations of PSO were aimed at the phenomenon of prematurity, but most of them added complexity to the paradigm and made PSO harder to be understood. In this paper, the evaluation of a particle’s position was investigated, and it was proposed that the evaluation strategy in standard PSO had two shortcomings, i.e. “two steps forward, and one step back” and “two steps back, and one step forward”. A novel evaluation strategy with the same convergence analysis as the standard PSO was presented, whereby each particle was evaluated in dimension-by-dimension order so as to step steadily toward the aimed position, and experiments showed that the novel strategy was very promising for multimodal and high-dimensional problems. A general evaluation strategy was defined, and experiments showed that the number of grouping could balance effectively the convergence speed and the best fitness so to get excellent performance.
