Multi-objective evolutionary algorithms rely on the use of variation operators as their basic mechanism to carry out the evolutionary
process. These operators are usually fixed and applied in the same way during algorithm execution, e.g., the mutation probability in genetic algorithms. This paper analyses whether a more dynamic approach combining different operators with variable application rate along the search process allows to improve the static classical behavior. This way, we explore
the combined use of three different operators (simulated binary crossover, differential evolution’s operator, and polynomial mutation) in
the NSGA-II algorithm. We have considered two strategies for selecting the operators: random and adaptive. The resulting variants have been
tested on a set of 19 complex problems, and our results indicate that both
schemes significantly improve the performance of the original NSGA-II
algorithm, achieving the random and adaptive variants the best overall
results in the bi- and three-objective considered problems, respectively.