本論文提出一種改良式粒子群最佳化演算法,其目的為改善傳統粒子群最佳化演算法在多目標最佳化問題的效能。演化的過程分成兩個階段:階段一,將問題的搜尋空間分割成數個子空間,並利用多個粒子群找到搜尋空間中大多數的區域最佳解,且在演化的過程中粒子們會在族群間移動;階段二,利用階段一所找到的數個區域最佳解組成一組新的粒子群,並繼續對整個空間作全域最佳解的搜尋。根據最後的實驗結果顯示改良式粒子群最佳化演算法在解決多數的多目標最佳化問題上有很好的結果。
This paper presents an improved particle swarm optimization which improved the efficiency on the multimodal optimization problems. The new algorithm has two stages: In the first stage, we split the problem’s search space into k sub-space, and then using k particle swarms to find the optimum in each sub-space, the local optimum in the original search space. During this stage, particles can move to different swarms. In the second stage, we organize the several local optimums finding in the first stage into a new swarm, and continue searching for the global optimum. Empirical examination of the evolution shows that the improved PSO has better efficiency than PSO.