Baldwin Effect based Particle Swarm Optimizer for Multimodal Optimization

Abstract

Particle Swarm Optimization (PSO) is an effective optimal technique. However, it often suffers from being trapped into local optima when solving complex multimodal optimizing problems due to its inefficient exploiting of feasible solution space. This paper proposes a Baldwin effect based learning particle swarm optimizer (BELPSO) to improve the performance of PSO when solving complex multimodal optimizing problems. This Baldwin effect based learning strategy utilizes the historical beneficial information to increase the potential search range and retains diversity of the particle population to discourage premature. On the other hand, the exemplars provided by Baldwin effect based learning strategy can flatten out the fitness landscape closing to optima and hence guide the search path towards optimal region. Experimental simulations show that BELPSO has a wider search range of feasible solution space than PSO. Furthermore, the performance comparison between BELPSO and amount of population based algorithms on sixteen well-known test problems shows that BELPSO has better performance in quality of solution.