Robust berth scheduling using machine learning for vessel arrival time prediction

Abstract In this work, the potentials of data-driven optimization for well-known berth allocation problem are studied. The aim robust scheduling is to derive conflict-free vessel assignments at quay a terminal, taking into account uncertainty regarding actual arrival times which may result from external influences as, e.g., cross wind and sea current. order achieve robustness, four different Machine Learning methods-from linear regression an artificial neural network-are employed time prediction in work. methods analysed evaluated with respect their forecast quality. calculation use so-called dynamic buffers (DTBs), derived AIS-based forecasts whose length depends on estimated reliability, model enhance robustness resulting schedules considerably, as shown extensive numerical study. Furthermore, results show that also rather simple approaches able reach high accuracy. does not only lead more solutions, but less waiting vessels hence enhanced service quality, can be by studying real data. Moreover, it turns out accuracy berthing schedules, measured deviation planned actually realisable exceeds all underlines usefulness DTB approach.