A Multimodal Multi-Objective Evolutionary Algorithm Based on Historical Archive

For multimodal multi-objective optimization problems, numerous algorithms have been developed and have achieved promising results. However, a common challenge is that many works do not make full use of historical information. Even if an excellent solution is found, it cannot be retained, which leads to accidental deletion or convergence diversity oscillation. This paper proposes a historical archive-based algorithm, named MMO_HA. The algorithm reintroduces discarded individuals during iterations via a historical archive (HA), leveraging a neighborhood-based method that considers the diversity of these individuals in both the decision and objective spaces. The HA is updated alongside the population and eventually merged with the optimal archive (EXA), providing a positive supplement to EXA. Finally, a selection process based on decision-space neighborhood strategies using objective-space crowding distances is applied to further enhance diversity. Experimental results on the 2019 CEC test suite demonstrate the effectiveness of the proposed method.