KIRM

Project goals

The aim is to increase efficiency and sustainable use of resources. Terminal-side processes and rail-based operations are predicted in order to minimize bottlenecks. Among other things, the focus is on reducing waiting times, empty runs and emissions, which should result in more efficient port operations.

In the future, a robust, AI-supported basis will be created that will generate sustainable capacity utilization and savings in the long term.

 

Objectives

KILOG builds on advanced AI forecasting models and Large Language Models (LLMs) to comprehensively optimize container processes in ports. The project uses state-of-the-art methods to forecast incoming and outgoing container flows via ship and rail, supplemented by high-frequency data. These are processed using a hybrid lakehouse structure so that a wide range of data types can be integrated.

This makes it possible to refine terminal-side processes in a targeted manner, while significantly reducing the use of resources.

KILOG extends classic time series models with LLM components for unstructured data. An agile approach integrates real-time information into a hybrid lakehouse architecture so that structured and unstructured sources can be evaluated together. Forecasting approaches are developed for ship, rail and terminal operations, trained using historical data and iteratively optimized. LLMs process messages, making predictions on container volumes and warehouse utilization even more precise. Dynamic adjustments react to changes in plans or delays. After extensive testing, integration into existing systems takes place. Interfaces feed optimized processes, enabling bottlenecks to be identified at an early stage and making port logistics more sustainable overall.

 

Tasks of the CML

- Participation in the development of use cases at the HHLA terminals

- Conception of various solution approaches for the developed use cases

- Development of various AI models for forecasting container movements

- Development of LLM-based approaches for the preparation of unstructured data

- Integration of real-time data and dynamic adjustment mechanisms into the solution approaches

- Development of mechanisms for self-learning AI models

- Contributing to the integration of the solution approaches into HHLA's existing system landscape

 

Project consortium

The consortium consists of HHLA AG (coordinator) and the Fraunhofer CML.

Funding framework

The project is funded over two years by the IHATEC II - Innovative Port Technologies II funding program of the Federal Ministry of Digital and Transport. The duration of the project is from 03/2025 to 02/2027.

Project volume: € 1,116,423.40 (72% of which is funded by the BMV)