Multi-Agent-based Decision Automation in Maritime Container Logistics
Decision-making concerning a container’s next transport step provides the opportunity for automation. Today, choosing a container’s transport vehicle, route, time, and price is a manual, human decision process. Fast changing conditions in the market, globally acting supply chains and permanently growing transport volumes have extensive effects on the planning and control of logistic processes. The dynamic and broad branching of transport chains leads to an increasing amount of possible transport passages and thus an increased amount of information. This complicates the information supply to central decision instances and impedes human decision processes that cannot take into account all options. In order to manage this problem, central systems (i.e. central decision), are not appropriate for the future. Thus, a new control paradigm for logistic processes, concentrating on dynamic decentralized autonomous decisions instead of central control, is required. This paradigm shifts central planning to many decentralized logistic objects, i.e. to containers and transport vehicles.
Existing automation approaches often focus on (1) optimizing complete transport routes with an a priori overall knowledge of all options and (2) coping with occurring emergencies resulting from traffic jams, broken cold chains in the context of temperature-sensitive and high-perishable goods, or using sensors within a container to initiate decision processes. For example, they create a scenario where transport vehicles and containers equipped with sensors are registered and permanently connected to a central service broker, a routing system, a traffic information system etc. This scenario focuses on (1) high-perishable goods in the containers and (2) predefined routes in a specific area.
The approach of this project is to choose a more general perspective, because we will take enhancements into account, which benefit all kinds of maritime containers and will not limit our research to reefer containers. Also, our approach focuses more on the container transshipment to investigate the negotiation and decision making in the context of further container transport and does not deal with sensory input during the transport, but rather with container and shipment specific data. We take all possible transport modes for 20 and 40ft containers into account (ship, train and truck). Even though the focus is on the container as a transport object, we also respect the case that a shipment contains multiple containers and vice versa.