A Systematic Comparison of Performance of Various Multiple Objective Metaheuristics Using a Common Set of Analytical Test Functions

摘要

Many multiple objective optimization algorithms have been described in the literature. Some of them use a “metaheuristic” (genetic algorithm, simulated an- nealing, tabu search and so on) that allow, in principle, to avoid getting trapped into a local minimum of an objective function. We feel that this approach can be advantageously extended to a large set of multiple objective optimization meth- ods. Moreover it is interesting to perform a systematic comparison between Per formances of various multiple objective metaheuristics. Such a comparison needs, on the one hand, to adopt a common set of test functions and, on the other hand, to use a common set of performance criteria. In this study, we propose to com- pare various metaheuristics associated with various multiple objective optimization methods (such as weighted sum of objective functions, goal programming, distance method and so on). These different couples are evaluated using a set of classical test functions. The set of test functions is chosen so as to represent most of the difficulties (multifrontallity, discontinuity, non-convexity and so on) that can be met in engineering when handling real multiple objective optimization problems.