A New Genetic Algorithm for Sequencing the Multiobjective Flowshop

Abstract

This chapter describes a new genetic algorithm that obtains Pareto optimal solutions faster than NSGA. When NSGA is applied to realistic multiobjective problems it is often seen that it lacks somewhat in both on-line performance (converging rapidly to good solutions) and off-line performance (ensuring superior quality of the final solutions). One major reason for this is that NSGA does not preserve the good solutions found from one generation to the next generation. Thus, good (near-optimal) solutions lost in one generation have only a probabilistic chance in NSGA to reappear in the future. Also, the number of final solutions on the Pareto optimal front in NSGA often remains relatively low even with good choice of parameters and even after many generations, unless large population sizes (> 250) are used.