1. Introduction
The global shipping industry faces urgent pressure to reduce pollution to the environment and decarbonize while maintaining economic viability.
With the IMO targeting net-zero emissions by 2050, transitions to alternative fuels like green ammonia, liquid hydrogen, and dual-fuel systems promise reductions of up to 90% in CO2, but introduce challenges: limited refuelling infrastructure, varying energy densities, and route dependencies that could inadvertently increase emissions if not optimized. This PhD project, funded through Sust.AI, harnesses cutting-edge AI and data analytics to address these issues, developing tools that predict vessel behaviours from imperfect real-world data and simulate sustainable pathways forward.
Hosted at the Department of Decision Analytics and Risk under the supervision of Prof. Patrick Beullens, a leading expert in logistics analytics and alternative shipping fuels (see references), and Dr James Stallwood, an expert in computer science and AI, the project leverages exclusive access to high-fidelity Automatic Identification System (AIS) data. AIS provides near-real-time vessel positions, speeds, and headings, but suffers from gaps, delays, and inaccuracies reported by captains—issues that undermine traditional forecasting. This research will advance probabilistic AI models to deliver reliable, uncertainty-aware predictions.
Core Innovations:
Trajectory Prediction Under Uncertainty: Employ Transformer-based deep learning (e.g., attention mechanisms for long-range dependencies) fused with Bayesian neural networks to forecast destinations and routes for tankers like VLCCs. This enables "what-if" scenarios: How might economic shifts (e.g., volatile oil markets) alter shipping patterns, helping operators select profit-maximizing ports while avoiding congestion?
Decision Optimization: Integrate predictions into Partially Observable Markov Decision Processes (POMDP) frameworks enhanced by reinforcement learning, allowing vessels to dynamically choose routes that balance long-term profits with sustainability constraints—e.g., detouring to hydrogen-compatible ports despite added distance.
Environmental Impact Forecasting: Simulate fleet-wide transitions using agent-based models, quantifying trade-offs for fuels like ammonia (high energy density but NOx risks) versus hydrogen (zero-emission but short range). Outputs will include emission heatmaps, infrastructure gap analyses, and policy recommendations, revealing how better movement understanding can amplify decarbonization benefits (e.g., 20-30% GHG cuts via optimized bunkering).
The successful candidate—a motivated PhD with a background in AI, data science, or environmental engineering—will work in a collaborative environment with industry partners, gaining hands-on experience with proprietary datasets and publishing in top relevant journals and presentation of their work at leading academic conferences. This project not only advances AI's role in handling "noisy" real-world data but also contributes to a greener maritime future, fostering innovations that could save billions in fuel costs and avert millions of tons of emissions annually. Applications are welcome from diverse candidates passionate about AI for planetary good.
References:
Beullens, P. 2025. Liquid Hydrogen Mobile Storage and Dispensing for Aircraft in the UK. University of Southampton, Deliverable to Innovate UK project 10076104, 204 pages.
Xu, K., Brito, M.P. and Beullens, P., 2025. Multi-criteria feature selection on maritime emission abatement alternatives. Research in Transportation Business & Management, 59, p.101288.
Svirschi, O., Beullens, P. and Arruda, E., Optimal Speed and Hedging Strategies for Tramp Shipping Operators in Volatile Freight Markets. Available at SSRN 5114763.
Song, Y., Beullens, P. and Hudson, D., Job Acceptance and Economic Travel Time of a Tramp Ship Under Risk. Available at SSRN 4821126.
Beullens, P., Ge, F. and Hudson, D., 2023. The economic ship speed under time charter contract—A cash flow approach. Transportation Research Part E: Logistics and Transportation Review, 170, p.102996.
Ge, F., Beullens, P. and Hudson, D., 2021. Optimal economic ship speeds, the chain effect, and future profit potential. Transportation Research Part B: Methodological, 147, pp.168-196.