The particle swarm algorithm

Abstract

Many algorithms are the result of biological inspiration and particle swarm optimization (PSO) is no exception. However, the PSO algorithm has slightly different end goals to the biological behavior that provides its inspiration and so needs to differ from its biological inspiration in some, perhaps non-biological, ways. PSO takes its inspiration from the flocking of birds and fish. In the real world, the flock needs to be compact for protection, and once food is found the flock should settle to feed. In the artificial particle swarm optimization the aim of the algorithm is to find an optimum solution to some problem, rather than the protection or food sought in the natural environment. For PSO the correct behavior once an optimum is found is not for all the particles in the swarm to converge on this, possibly local, optimum as the goal is to check many optima in the hope of finding the global optimum. Instead of converging, once an optimum has been found, it should be noted and the particles should immediately disperse to look for another, perhaps better, optimum. In Nature the time will come when a swarm that is feeding has consumed the food so that the place is no longer optimal: if swarming for protection the threat may change or even disappear completely. Then the swarm will again set out. Such extended periods of convergence serve no useful purpose so far as an artificial swarm is concerned. © 2008 Springer-Verlag Berlin Heidelberg.