Particle swarm optimization (PSO) is a well-known swarm intelligence algorithm inspired by the foraging behavior of bird flocking. PSO has been widely used in many optimization and engineering problems due to its simplicity and efficiency, even though there still exist some disadvantages. The standard PSO often suffers with premature convergence or slow convergence when the optimization problem is multimodal or high-dimensional. To overcome these drawbacks, an ecosystem PSO (ESPSO) inspired by the characteristic that a natural ecosystem can excellently keep the biological diversity and make the whole ecosystem be in a dynamic balance is presented in this paper. ESPSO not only prevents the algorithm trapping into local optima but also balances the exploration and exploitation in both unimodal and multimodal problems as compared to other PSO variants. Twenty benchmark functions including unimodal functions and multimodal nonlinear functions are used to test the searching ability of ESPSO. Experimental results show that ESPSO considerably improves the searching accuracy, the algorithm reliability and the searching efficiency in comparison with other six well-known PSO variants and four evolutionary algorithms. Moreover, ESPSO was successfully applied to the antenna array pattern synthesis design and gained satisfactory results.
Read full abstract