Evolutionary algorithms for multi-objective optimization: Performance assessments and comparisons

Evolutionary techniques for multi-objective (MO) optimization are currently gaining significant attention from researchers in various fields due to their effectiveness and robustness in searching for a set of trade-off solutions. Unlike conventional methods that aggregate multiple attributes to form a composite scalar objective function, evolutionary algorithms with modified reproduction schemes for MO optimization are capable of treating each objective component separately and lead the search in discovering the global Pareto-optimal front. The rapid advances of multi-objective evolutionary algorithms, however, poses the difficulty of keeping track of the developments in this field as well as selecting an existing approach that best suits the optimization problem in-hand. This paper thus provides a survey on various evolutionary methods for MO optimization. Many well-known multi-objective evolutionary algorithms have been experimented with and compared extensively on four benchmark problems with different MO optimization difficulties. Besides considering the usual performance measures in MO optimization, e.g., the spread across the Pareto-optimal front and the ability to attain the global trade-offs, the paper also presents a few metrics to examine the strength and weakness of each evolutionary approach both quantitatively and qualitatively. Simulation results for the comparisons are analyzed, summarized and commented.