Optimizing Roll-on/Roll-off Stowage Planning: A Study of Artificial Intelligence and Operations Research Methods

Roll-on/roll-off (RORO) shipping plays an essential part in global logistics. This PhD thesis comprises three chapters addressing several approaches to optimizing the dynamics of RORO stowage planning using operations research and machine learning methods.

The first chapter introduces how operations research techniques can be applied to solving the RORO stowage planning problem with dynamic cargo arrival times. Using mathematical programming and metaheuristics, the chapter concludes that the proposed method can generate feasible stowage plans. Yet, more research is required to enhance the robustness of the approach.

The second chapter introduces how machine learning techniques can be applied to RORO stowage planning with uncertain cargo arrival times. The problem can be solved to near optimality using deep reinforcement learning. This further encourages reinforcement learning approaches in combinatorial optimization with uncertainty.

The third chapter introduces how machine learning techniques can be incorporated into operations research techniques and applied to RORO stowage planning. This interdisciplinary methodology is evaluated by analyzing the RORO stowage planning problem with multiple discharge ports, heterogeneous cargo sizes, and simple stability. The chapter illustrates that the proposed methodology can create good stowage plans, yet more research is necessary to improve its effectiveness.

This thesis contributes with novel approaches to the research field of RORO stowage planning. The thesis expands the knowledge about how OR and ML methodologies, separately and combined, can be applied to optimize RORO stowage planning to increase the industry’s competitiveness, reduce consumption, and combat climate change.