AI-Driven Digital Transformation of Shipping and Haulage Operations

Repositioning of empty containers to enable efficient inter-modal transportation

Within the shipping/haulage industry the efficient movement and positioning of empty containers for fulfilling import/export orders is critical to meet customer’s demand. However, this comes at considerable costs due to transportation and storage involved in over-land haulage maritime transportation

Business challenge 

Traditionally containers are managed based on human-driven planning utilising data from the previous day, week, or month. This can be inefficient due to:

• Uneven supply and demand for empty containers 
• Uncertainty in the future supply and demand chain
• Complex business constraints
• Complicated Logistic Networks (ocean, road, rail)
• Dynamic factors, e.g., special day/event, emerging market changes, and unstable policies

The technical challenge

• Understand the environment topology (percentage of import/export at the port, different types/size container, full/empty proportion, storage, and transport capacity) and build an event-driven data pipeline.
• The data pipeline can be configured to use synthetic and real data for simulating these scenarios.
• Creating a Machine Learning (ML) model: Developing a multi agent-based mechanism to monitor the supply and demand of empty containers; Optimising the movement of the container at a reduced cost flow.

The solution

We developed a prototype "Empty Container Repositioning (ECR)" system for MSC that can simulate order data in challenging environmental topologies and recommend ECR actions using an AI model. The model was based on using Graph Neural Networks, Transformer/Self-Attention, Deep Learning, Multi-Agent Reinforcement learning (MARL).

Through the simulation environment the models were able to improve the import/export orders fulfilment percentage and reduce the unnecessary empty moves, improving the availability of containers at a location when needed.