An evolutionary Game-Theoretical approach to Particle Swarm Optimisation

Abstract

This work merges ideas from two very different areas: Particle Swarm Optimisation and Evolutionary Game Theory. In particular, we are looking to integrate strategies from the Prisoner Dilemma, namely cooperate and defect, into the Particle Swarm Optimisation algorithm. These strategies represent different methods to evaluate each particle's next position. At each iteration, a particle chooses to use one or the other strategy according to the outcome at the previous iteration (variation in its fitness). We compare some variations of the newly introduced algorithm with the standard Particle Swarm Optimiser on five benchmark problems. © 2008 Springer-Verlag Berlin Heidelberg.