Evaluation of sequential, multi-objective, and parallel interactive genetic algorithms for multi-objective optimization problems

Abstract

We propose a sequential interactive genetic algorithm (IGA), multi-objective IGA and parallel IGA, and evaluate them with both simulated and real users. Combining human evaluation with an optimization system for engineering design enables us to embed domain- specific knowledge that is frequently hard to describe, i.e. subjective criteria, and design preferences. We introduce a new IGA technique to extend the previously introduced sequential single objective GA and multi-objective GA, viz. parallel IGA. Experimental evaluation of three algorithms with a multi-objective manufacturing plant layout design task shows that the multi-objective IGA and the parallel IGA clearly provide better results than the sequential IGA, and that the multi-objective IGA gives the most diverse results and fastest convergence to a stable set of qualitatively optimum solutions, although the parallel IGA provides the best quantitative fitness convergence.