Abstract:

In order to improve the ability to escape from local optima, we present an improved particle swarm optimizer based on dynamic population and comprehensive learning (DCPSO for short). In DCPSO, the swarm population growing and declining strategies are introduced to increase the swarm diversity, further improve the ability to escape from local optima; a comprehensive learning strategy also is used to improve the probability of flying to the global best position. In the benchmark function, the results demonstrate good performance of the DCPSO algorithm in solving complex multimodal problems when compared with other PSO variants. In the optimization design for the box grider of portal gantry, the experimental results show that the DCPSO algorithm can achieve better solutions that other PSOs.

Key words: dynamic population;comprehensive learning;particle swarm optimizer