EGU25-19197, updated on 15 Mar 2025
https://doi.org/10.5194/egusphere-egu25-19197
EGU General Assembly 2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
Agentic AI for ship routing
Fearghal O'Donncha1, Abigail Langbridge2, Alexander Timms3, Antonis Antonopoulos4, Antonis Mygiakis4, and Eleni Voulgari4
Fearghal O'Donncha et al.
  • 1IBM Research Europe - Ireland
  • 2Dyson School of Design Engineering, Imperial College London, UK
  • 3Department of Bioengineering, Imperial College London, UK
  • 4Konnecta, Kifissia, Athens, Greece

The complexity of the shipping industry, with its dynamic operational drivers and diverse data sources, presents significant scalability challenges for digital twins. Agentic Large Language Models (LLMs), augmented with external tools, offer a promising solution to streamline operations and improve decision-making. By leveraging pre-trained knowledge and reasoning capabilities, these LLMs can autonomously select the most relevant tools and data streams, facilitating real-time decision-making that optimizes ship routes, fuel consumption, and operational efficiency.

In this demonstration, we explore how agentic LLMs can enhance the scalability, flexibility, and efficiency of digital twins in shipping by optimising route planning with consideration for weather conditions, fuel consumption, and speed. By integrating weather data and analysing trade-offs between fuel consumption, speed, and routing choices, the system enables more effective decision-making to balance operational goals with environmental considerations. This approach facilitates a deeper understanding of how shipping operations can be adjusted for reduced emissions and improved fuel efficiency while considering the complexities of real-world constraints.

We showcase how this agentic digital twin solution supports more efficient route optimisation, ultimately contributing to the shipping industry’s transition to low-carbon fuels and reduced environmental impacts. This interactive system demonstrates the potential of agentic LLMs to reduce operational complexity and improve the practical application of digital twins in real-world settings.

How to cite: O'Donncha, F., Langbridge, A., Timms, A., Antonopoulos, A., Mygiakis, A., and Voulgari, E.: Agentic AI for ship routing, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19197, https://doi.org/10.5194/egusphere-egu25-19197, 2025.