Deep Learning for Solving Loading, Packing, Routing, and Scheduling Problems

Machine learning encompasses several approaches, including reinforcement and deep learning. The main objective of RL is to maximize the rewards obtained by an agent through its actions in accordance with the state of the environment. Deep learning has proved to be efficient in solving complex optimization problems. In this study, we investigate the use of deep learning (DL) to solve combinatorial optimization problems related to scheduling, packing, loading, and routing. A systematic literature review is run in well-known scientific databases such as Scopus, Web of Science, and IEEE Xplore. Once the papers were screened, a set of 25 papers that pertain to the scope of this study were retained. All these papers are reviewed and classified in detail. The studies selected show that increasing attention is being given to DL to solve combinatorial optimization problems over the years. Precisely, the Q-learning and policy gradients are the most used algorithms, and the scheduling and loading problems are, respectively, the most and the least handled.