Real-time multi-resource jointed scheduling of container terminals with uncertainties using a reinforcement learning approach

Abstract

The automated container terminal is a large operation system, the efficiency of it is highly dependent on the jointed scheduling of container handling equipment. The scheduling of automated container terminals is a difficult challenge, especially considering the terminal’s uncertainties. In order to meet this obstacle, this study proposes an approach based on reinforcement learning (RL) to promote the jointed scheduling of quay cranes (QCs) and automated guided vehicles (AGVs), the uncertainties are considered with the goal of reducing the ship’s makespan. A simulator for the real-time multi-resource scheduling is built with Qingdao port automated terminal data. Empirical studies show that the proposed approach can further improve the efficiency of automated container terminals compared with several existing scheduling methods.