Abstract: Hyper-heuristics algorithms seek to efficiently solve the combinatorial optimization problems. It is expected to handle classes of problems rather than solving just one problem. This paper presents a systematic review of the research progress of hyper-heuristic algorithms. Based on a brief introduction to hyper-heuristic definition, structure and classification, the heuristic selection and heuristic generation approaches of hyper-heuristics are summarized. The techniques such as the machine learning methods in choosing low-level heuristics, the move acceptance criterions for neighborhood solutions, the generation of low-level heuristics, and the algorithm framework are discussed in details. Finally, the limitations and further research directions of hyper-heuristics algorithms are discussed.

Key words: hyper-heuristics algorithm, selection, generation

摘要： 超启发算法是一类新兴的优化方法，通过机器学习、算法选择、算法生成等技术求解组合优化等问题，具备跨问题领域求解的能力。针对超启发算法研究进展进行综述和讨论。首先，梳理超启发算法的定义、结构、特点和分类；其次，归纳选择式超启发算法和生成式超启发算法的研究进展及相关技术，包括选择低层启发式算法采用的学习方法，迭代计算中的移动接受策略，低层启发式算法的生成方法；最后，讨论现有超启发算法研究中存在的不足及未来的研究方向。

关键词: 超启发算法, 选择式, 生成式