Reinforcement learning applications to machine scheduling problems: a comprehensive literature review

Reinforcement learning (RL) is one of the most remarkable branches of machine learning and attracts the attention of researchers from numerous fields. Especially in recent years, the RL methods have been applied to machine scheduling problems and are among the top five most encouraging methods for scheduling literature. Therefore, in this study, a comprehensive literature review about RL methods applications to machine scheduling problems was conducted. In this regard, Scopus and Web of Science databases were searched very inclusively using the proper keywords. As a result of the comprehensive research, 80 papers were found, published between 1995 and 2020. These papers were analyzed considering different aspects of the problem such as applied algorithms, machine environments, job and machine characteristics, objectives, benchmark methods, and a detailed classification scheme was constructed. Job shop scheduling, unrelated parallel machine scheduling, and single machine scheduling problems were found as the most studied problem type. The main contributions of the study are to examine essential aspects of reinforcement learning in machine scheduling problems, identify the most frequently investigated problem types, objectives, and constraints, and reveal the deficiencies and promising areas in the related literature. This study can help researchers who wish to study in this field through the comprehensive analysis of the related literature.