Modeling the Efficiency of a Port Community System as an Agent-based Process

We present an agent-based method which makes use of reinforcement learning in order to estimate the efficiency of a Port Community System. We have evaluated the method using two weeks of observations of import containers at the Port of Brisbane as a case study. Three scenarios are examined. The first scenario evaluates the observed container delivery by individual shipping lines and estimates the consignments allocated to the various road carriers based on optimizing the individual shipper's total logistics cost. The second scenario implies that, in the optimum case, all agents (shipping lines and road carriers) communicate and cooperate through a single portal. The objective of cooperation is in sharing vehicles and creating tours to deliver shipments to several importers in order to reduce total logistics costs, while physical and time window constraints are also considered. The third scenario allows for some agents to occasionally decide to act based on individual costs instead of total combined logistics costs. The results of this study indicate an increase in the efficiency of the whole logistics process through cooperation, and the study provides a prototype of a Port Community System to support logistics decisions.