TOPOLOGY AND PARAMETER SELECTION OF PARTICLE SWARM OPTIMIZATION- A HEURISTICS STUDY

Swarm intelligence deals with studying collective behavior of intelligent agents, capable of interacting locally with neighbors so as to create coherent global functional patterns. Particle Swarm optimization algorithm is one important member of the swarm intelligence family. This algorithm employs a number of agents, called particles, the position and velocity of which are adapted over time with an ultimate objective to search the optima in a given search landscape. The motion of individual particle is controlled by its inertia and two attractive forces respectively along the historical and the global best (gbest) positions of the particle. The historical and the global best positions are adapted over iterations in the optimization algorithm. The paper provides an extensive review of the Particle Swarm Optimization algorithm. It outlines the basic algorithm, called gbest PSO introduced above and its extension called l-best PSO, where the search is explored locally by replacing the gbest with local best (lbest) position in the neighborhood of individual particle. Next the paper examines the strategy for parameter selection of the algorithm. Fundamental results on convergence of the algorithm are briefly outlined. Several variants of PSO for solving binary, integer, multi-objective, constrained, parallelization, large scale and dynamic optimization problems are discussed in brief. The review ends with a discussion on application of the PSO algorithms in different optimization problems.