Robustness Analysis in Evolutionary Multi-Objective Optimization

This paper presents two approaches to robustness analysis in multi-objective optimization problems, in which the model data (coefficients of objective functions, coefficients of constraints, bounds of decision variables, etc.) are subject to small perturbations, with respect to a ”nominal” set of coefficients of the model data. In these approaches, the concept of degree of robustness is incorporated into an evolutionary algorithm, being operationalized in the computation of the fitness value assigned to solutions. Non-dominated solutions are classified according to their degree of robustness. The information on the degree of robustness of solutions is provided to support the decision maker in the selection of a robust compromise solution.