The Research And Optimization Of Imperialist Competitive Algorithm

Swarm intelligence optimization algorithm has some special features different from those traditional optimization algorithms, such as global search ability,strong commonality and good parallelism and robustness. It is suitable for solving complex optimization problem, so it has been a focus in recent decades. Imperialist Competitive Algorithm (ICA) is a new algorithm for optimization which is inspired by the imperialistic competition. Like other intelligence optimization algorithms such as particle swarm optimization (PSO) and ant colony optimization (ACO), ICA belongs to a random optimization algorithm based on population search. Unlike some optimization algorithms based on biological behavior, ICA originated in a simulation of social behavior, therefore it is gradually concerned by scholars. However, there are some common problems in ICA and other intelligence optimization algorithms:convergence rate, optimization accuracy, local optimum and so on. Several kinds of improvement strategies were proposed in this paper, and the major contents include the following three points:(1) Because the number of empire reduces progressively in the iteration process, ICA’s population diversity declines. This situation makes ICA easy stuck into a local optimum when solves high-dimensional multimodal optimization problems. To overcome this shortcoming, an empire splitting strategy was proposed which obtains significant improvements when dealing with high-dimensional optimization problems. The experimental results on several benchmark functions validate the good property of the proposed algorithm.It indicates that proper splitting strategy is effective for improving the performance of ICA.(2) In the traditional ICA, colony revolution will lead to low precision because the operator may make the strong colony lost. To overcome this shortcoming, a differential evolution operator is introduced, which makes use of the interaction among colonies to produce new colonies. The operator will enhance population diversity and keep the excellent individuals at the same time. Furthermore, on account of strengthening the interaction among empires, a clone evolution operator is introduced, which includes the following steps:clonal reproduction of the stronger countries;high frequency variation and random crossover of clonal populations;the stronger countries take place of the weaker ones. The operator can guide the search for global optimal efficiently. The proposed methods are applied to six benchmark function and six typical complex function optimization problems, and the comparison of the performance of the proposed methods with other ICA algorithms is experimented. The results indicate that the proposed methods can significantly speed up the convergence and improve the precision and stability.(3) Discussing the practical application in solving traveling salesman problem (TSP) with ICA. Through coding the continuous ICA, a kind of discrete ICA (DICA) was proposed, which applies to dealing with TSP problems. The experimental results on ten TSP instances in TSPLIB validate the correctness and applicability of DICA.