A survey of surrogate-assisted evolutionary algorithms for expensive optimization

In practical engineering applications, many problems involve high computational costs in evaluating the objective function during optimization. Traditional optimization algorithms may require a large number of evaluations to find the optimal solution, which leads to large consumption of computational resources. In recent years, surrogate-assisted evolutionary algorithms (SAEAs) have received increasing attention in solving computationally expensive optimization problems (EOPs). This paper provides a review of research on surrogate-assisted evolutionary algorithms. Firstly, it introduces the characteristics and challenges of expensive optimization problems. Secondly, it introduces the framework of SAEAs and the representative single-objective and multi-objective expensive optimization algorithms. Then, it presents methods for surrogate model construction and model management strategy, summarizes relevant literature, and analyzes the characteristics of different methods. Finally, it concludes existing challenges and future research directions in this topic. Through a comprehensive review and analysis of surrogate-assisted evolutionary algorithms, this paper provides essential references and guidance for further research.