Dynamic optimization with an improved θ-PSO based on memory recall

A comparative study of θ-PSO and its improved model with partial particles randomization strategy on their abilities of tracking extrema in dynamic environments was carried out in our earlier work. And the results shown that θ-PSO has better performance in dynamic optimization than standard PSO. In this paper, an improved θ-PSO with memory recall and varying scale randomization strategy (θ-PSO-MR) is put forward. The eligible memory particles are recalled when the landscape changes. And the vary scale randomization is introduced through the evolution to maintain the swarm diversity. The offline error in the non-trivial multimodal dynamic functions MPB indicates that this improved θ-PSO deals well with the complex dynamic tracking and optimization. And in some cases, θ-PSO-MR outperforms θ-PSO-Rn for the introduction of memory recall.