A two-stage particle swarm optimizer

This paper presents a variant of particle swarm optimizers (PSOs), called the two-stage particle swarm optimizer (TSPSO). TSPSO performs a gross searching algorithm at the first stage, and switches to a fine-grained searching algorithm if it is stagnated at the first stage. A switching criterion was proposed, and a new fine-grained searching algorithm was devised to work at the second stage of TSPSO. For the first stage, Fully Informed PSO (FIPS) with U-square topology was adopted. At the second stage, the fine-grained searching algorithm has very good performance on complex multimodal functions such as Rastrigin and Schwefel functions. The switching behavior makes TSPSO adaptive to the problems to be solved. Experimental results show that TSPSO has very good performance on both unimodal and multimodal functions compared with six other variants of PSO. Especially on complex multimodal functions, TSPSO´s performance is even better than the most state of art PSOs such as CLPS and CPSO-H.