A new adaptive inertia weight strategy in particle swarm optimization

Abstract

According to the principle of mechanics, a new adaptive inertia weight strategy is proposed. The strategy depends on particle's search states including its location and velocity instead of iteration times. Based on the proposed strategy, an inertia weight function is designed, which is continuous in real domain, thus it's easy to be implemented and the computation cost is low. Experiments on three benchmark functions, comparison between convergence speed, the ability to search the global solution of the linear decreasing strategy (LPOS) and the proposed strategy are done. The experimental results are also analyzed in detail.