Continuous Trait-Based Particle Swarm Optimisation (CTB-PSO)

Abstract

In natural flocks, individuals are often of the same species, but there exists considerable variation in the traits possessed by each individual. In much the same way as humans display varied levels of aggression, gregariousness and inquisitiveness, so do the animals on which PSO is based [1]. Recent research has shown that this disparity of behaviour is very important in the ability of the flock to solve problems effectively, which might have profound implications for PSO. One of the key aspects is that although certain behaviour types (e.g. more adventurous individuals) might individually be better at problem solving; selecting for a group that all have adventurous traits has been shown to reduce the performance of the flock as a whole [1]. Therefore a flock that has a variety of behaviours leads to better performance in natural systems and it is this that motivates the work here. This paper explores a variant of PSO known as Continuous Trait-Based PSO (CTB-PSO) where individuals within a swarm have traits based on a continuous scale as opposed to discrete behaviour groupings.