Abstract—
The Particle Swarm Optimization (PSO) algorithm,as one of the latest algorithms inspired from the nature, wasintroduced in the mid 1990s and since then, it has been utilizedas an optimization tool in various applications, ranging frombiological and medical applications to computer graphics andmusic composition. In this paper, following a brief introductionto the PSO algorithm, the chronology of its evolution ispresented and all major PSO-based methods arecomprehensively surveyed. Next, these methods are studiedseparately and their important factors and parameters aresummarized in a comparative table. In addition, a newtaxonomy of PSO-based methods is presented. It is the purposeof this paper is to present an overview of previous and presentconditions of the PSO algorithm as well as its opportunities andchallenges. Accordingly, the history, various methods, andtaxonomy of this algorithm are discussed and its differentapplications together with an analysis of these applications areevaluated.Index Terms – Heuristic Optimization, Particles SwarmOptimization (PSO), Taxonomy, Applications.
I. INTRODUCTION
Since conventional computing algorithms are not capableof solving real-world problems because of sometimes havingan inflexible structure mainly due to incomplete or noisy dataand some multi-dimensional problems, Natural computingparadigms seem to be a suitable replacement in solving suchproblems. These paradigms consist of simple elements thatcan solve complicated problems of the real world whenworking together. It should be mentioned that the maindrawback of such paradigms are their indefinite nature andpresenting an approximate solution. In general, Naturalcomputing paradigms can be divided into three categories: 1)Epigenesis 2) Phylogeny 3) Ontogeny.The Epigenesis group is related to a situation in which wewould like to develop an intricate structure and to do so, it isnecessary to perform a tentative learning. A clear example ofthis category is Artificial Neural Network (ANN) whereinhuman's brain is simulated as a complex system.The phylogeny group is related to EA algorithms. In thealgorithms related to this category, there is a competition among agents on survival of the fittest. Algorithms related tothis group include Evolutionary Programming (EP), GeneticProgramming (GP), and Differential Evolutionary (DE).The Ontogeny group is associated with the algorithms inwhich the adaptation of a special organism to its environmentis happened. The algorithms like PSO and GeneticAlgorithms (GA) are of this type and in fact, they have acooperative nature in comparison with other types [16]. Theadvantages of above-mentioned categories can be noted astheir ability to be developed for various applications and notneeding the previous knowledge of the problem space. Theirdrawbacks include no guarantee in finding an optimumsolution and high computational costs in completing FitnessFunction (F.F.) in intensive iterations. Among theaforementioned paradigms, the PSO algorithm seems to be anattractive one to study since it has a simple but efficientnature added to being novel. It can even be a substitution forother basic and important evolutionary algorithms.The most important similarity between these paradigmsand the GA is in having the seam interactive population. Thisalgorithm, compared to GA, has a faster speed in finding thesolutions close to the optimum and it is faster than GA inpremature convergence
II. DESCRIPTION OF PSO
Kennedy and Eberhart [31], considering the behavior ofswarms in the nature, such as birds, fish, etc. developed thePSO algorithm. The PSO has particles driven from naturalswarms with communications based on evolutionarycomputations. PSO combines self-experiences with socialexperiences. In this Algorithm, a candidate solution ispresented as a particle. It uses a collection of flying particles(changing solutions) in a search area (current and possiblesolutions) as well as the movement towards a promising areain order to get to a global optimum.

Download full report
http://ijctepapers/80-g580-502.pdf