Adapting Particle Swarm Optimization in Dynamic and Noisy Environments

Abstract

The optimisation in dynamic and noisy environments brings closer real-world optimisation. One interesting proposal to adapt the PSO for working in dynamic and noisy environments was the incorporation of an evaporation mechanism. The evaporation mechanism avoids the detection of environment changes, providing a continuous adaptation to the environment changes and reducing the effect when the fitness function is subject to noise. However, its performance decreases when the fitness function is not subjected to noise (with respect to methods that use environment change detection). In this paper we propose a new dynamic evaporation policy to adapt the PSO algorithm to dynamic and noisy environments. Our approach improves the performance when the fitness function is dynamic and not subject to noise. It also keeps a similar performance when the fitness function is subject to noise.