GMM-based online optimization for container stacking in port container terminals

Abstract

Online optimization has a limitation in that it creates a policy that is unrelated to the actual data by not char-acterizing the problem data uncertainty. In this study, to overcome the limitations of the vertical stacking policy which is a conventional online optimization approach utilized in the container stacking problem (CSP), we propose a GMM-based online optimization. The container weight is classified into data-driven weight classes based on the Gaussian mixture model (GMM), and our stacking policy is updated in response to problem data. When comparing the weight variance of the other existing stacking policies with the proposed stacking policy, the proposed stacking policy showed smaller values of weight variance on average. Based on this study, con-tainers can be stacked to facilitate flexible responses to various situations with uncertainties and reduce the time taken for container relocation movements.