Adaptability of a discrete PSO algorithm applied to the Traveling Salesman Problem with fuzzy data

Imperfection is a common characteristic of information nowadays. For example, in everyday life, decisions have to be made based on information that is incomplete, inconsistent, and/or uncertain. This inexactness makes the decision making a challenging task. This paper investigates the behavior of a well-known optimization method, Particle Swarm Optimization (PSO), when solving a fuzzy problem. The discrete PSO implementation is studied on a Traveling Salesman Problem (TSP) variant, designed to model the uncertain environmental influences. The experiments investigate several symmetric TSP instances and their fuzzy variants in order to study the impact of uncertain information in the quality of the results provided by PSO. The fuzzy variants were generated using a two-dimensional degree of fuzziness, which is proportional to the number of nodes of the instance. In addition, the amplitude of the uncertainty can be set at running time, so the degree of fuzziness used here is a systematic perturbation, providing similar effects on all studied TSP instances. The experimental results reveal that the PSO algorithm can handle uncertainty in data by showing good adaptability based on the used TSP benchmark set.