Studies on migration strategies of multiple population parallel particle swarm optimization

In order to solve more complex application problems, parallel particle swarm optimization (PPSO) was proposed in recent years and attracted more and more interests from researchers. Multipopulation PPSO is one of important models, it needs less processors and its communication cost is lower than fine-grained PPSO. When to design multipopulation PPSO, several additional parameters for parallelization should be considered first. They are topological structure, migration strategy, migration interval, migration rate, etc. Among them, migration strategy is an important parameter that heavily affects the performance of multipopulation PPSO. However, there is few work on this important parameter. In this paper, we proposed 8 migration strategies for multipopulation PPSO. Compared with most used One-To-Migrate strategies on 36 commonly used test functions, we found that both strategies BW and BWM are more efficient for high dimensionality problem, while on low dimensionality functions One-To-Migrate strategies are more effective. And what is better than our expected is that the strategy RR is most effective on some functions.