Integrating Particle Swarm Optimization with Stochastic Point Location method in noisy environment

Abstract

Particle Swarm Optimization (PSO) deteriorates when facing a high-noise environment. To address this issue, one popular mechanism is the resampling method that is based on re-evaluations to find the true fitness value. However, the budget for re-evaluations in PSO is limited. In this paper, we intend to integrate a Stochastic Point Location (SPL) method into PSO to alleviate the impacts of noise on the evaluation of true fitness. SPL deals with the problem of a learning mechanism locating a target point on the line in noisy environment. Up to now, Adaptive Step Searching is the fastest algorithm in solving the SPL problem and shows great anti-noise performance. This paper investigates two effective hybrid PSO approaches, by integrating PSO and PSO-Equal Resampling with Adaptive Step Searching. The simulation results and comparisons on 20 large-scale benchmark optimization functions in noisy environments demonstrate the superiority of the proposed approaches in terms of optimization accuracy and convergence rate.