Seeking optimal solutions in complex, multi-objective problems? This paper introduces the Pareto Archived Evolution Strategy (PAES), a simple evolution scheme designed to generate diverse solutions within the Pareto optimal set. PAES employs local search with a reference archive to estimate the dominance ranking of solution vectors. The study compares several PAES variants, including (1+λ) and (μ+λ) adaptations, against other algorithms such as the Niched Pareto Genetic Algorithm and the Nondominated Sorting Genetic Algorithm, using a set of six diverse test functions. Results are analyzed using techniques that transform attainment surfaces into univariate distributions for statistical comparison. Evidence from the tests demonstrates that PAES consistently performs well across various multiobjective optimization tasks. Providing a baseline approach, the research demonstrates PAES' effectiveness and competitiveness, particularly in scenarios where local search is advantageous. This study offers valuable insights for optimization strategies and algorithmic design.
Published in Evolutionary Computation, this paper contributes directly to the journal's focus on evolutionary algorithms and their applications. By introducing and evaluating a new evolution strategy, the research aligns with the journal's interest in advancing optimization techniques. The comparative analysis with existing methods further enhances the paper's relevance to the evolutionary computation community.