Options
A robust berth allocation optimization procedure based on machine learning
Publikationstyp
Conference Presentation
Publikationsdatum
2021-09-22
Sprache
English
Author
Citation
Hamburg International Conference of Logistics (HICL) 2021
Contribution to Conference
PURPOSE
In berth allocation planning, container vessels are to be assigned to berthing locations and times at the quay of a container terminal. Terminal operators often aim to provide the best possible service quality to the shipping companies, i.e. especially short waiting times. However, the actual arrival times of vessels are uncertain due to external influences, which impedes the planning and may lead to conflicts with respect to scheduled berths.
METHODOLOGY
Machine Learning techniques are applied to enable the determination of patterns in AIS data and hence to develop forecasts of the ships’ arrival times. With a robust optimization approach based on Dynamic Time Buffers, uncertainty is proactively considered in the planning phase, resulting in more robust berthing schedules.
FINDINGS
The results of this new approach are evaluated from an ex post perspective using real ship data and actual ship arrival times. It is shown by a numerical study that the average number of conflicts can be reduced significantly by this approach and that the new concept improves the schedules’ robustness.
ORIGINALITY
A novel concept of Dynamic Time Buffers is presented where the assigned buffer time depends on the level of uncertainty in the vessel’s arrival time and which leads to improved berthing plans.
In berth allocation planning, container vessels are to be assigned to berthing locations and times at the quay of a container terminal. Terminal operators often aim to provide the best possible service quality to the shipping companies, i.e. especially short waiting times. However, the actual arrival times of vessels are uncertain due to external influences, which impedes the planning and may lead to conflicts with respect to scheduled berths.
METHODOLOGY
Machine Learning techniques are applied to enable the determination of patterns in AIS data and hence to develop forecasts of the ships’ arrival times. With a robust optimization approach based on Dynamic Time Buffers, uncertainty is proactively considered in the planning phase, resulting in more robust berthing schedules.
FINDINGS
The results of this new approach are evaluated from an ex post perspective using real ship data and actual ship arrival times. It is shown by a numerical study that the average number of conflicts can be reduced significantly by this approach and that the new concept improves the schedules’ robustness.
ORIGINALITY
A novel concept of Dynamic Time Buffers is presented where the assigned buffer time depends on the level of uncertainty in the vessel’s arrival time and which leads to improved berthing plans.
Schlagworte
MLE@TUHH
DDC Class
330: Wirtschaft