6. Conclusion and future work
Both the memory scheme and the multi-swarm strategy are effective techniques for dynamic optimization problems. However, their hybrids have scarcely been studied. In most cases, employing the traditional population updating methods to enhance the performance of the multi-swarm algorithms by the memory may not yield ideal results. Aimed at the two aforementioned problems, a new swarm updating method is proposed in this paper and embedded into the typical species-based algorithm SPSO. The number of replaced particles, instead of being predefined, is set adaptively according to the number of species. This may alleviate the impact generated by the unsuitable assignment. To not substantially destroy the exploitation capability of each species, no more than one particle is replaced by the memory in each species. The retrieved memory particles are grouped into four categories and processed by different strategies. To demonstrate the validity and feasibility of the new method, the proposed algorithm MSPSO is compared with SPSO and its variants on both the MPB and CMPB. Moreover, MSPSO is tested on the DRPBG and compared with other algorithms. The experimental results show that MSPSO is competitive. The effect of the memory size on the performance of the algorithm is tested in different dynamic scenarios. The experimental results indicate that the optimal setting of the memory size is largely influenced by the number of peaks in the search space. If there are few peaks, a small memory size would be favorable.When the number of peaks becomes larger, a large memory size would be better. How to set the memory size adaptively is a topic of our future work.
