Neural network assisted branch and bound algorithm for dynamic berth allocation problems

One of the key challenges in maritime operations at container terminals is the need to improve or optimize berth operation schedules, thus allowing terminal operators to maximize the efficiency of quay usage. Given a set of vessels and a set of berths, the goal of the dynamic berth allocation proble...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:European journal of operational research Jg. 319; H. 2; S. 531 - 542
Hauptverfasser: Korekane, Shinya, Nishi, Tatsushi, Tierney, Kevin, Liu, Ziang
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.12.2024
Schlagworte:
Branch and bound
Dynamic berth allocation problem
Neural network
Branch and bound
Dynamic berth allocation problem
Neural network
ISSN:0377-2217
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:One of the key challenges in maritime operations at container terminals is the need to improve or optimize berth operation schedules, thus allowing terminal operators to maximize the efficiency of quay usage. Given a set of vessels and a set of berths, the goal of the dynamic berth allocation problem is to determine the allocation of each vessel to a berth and the berthing time that minimizes the total service time. This problem can be solved using exact solution methods such as branch and bound (BB) algorithms or heuristic methods, however, exact methods do not scale to large-scale terminal operations. To this end, this paper proposes a BB algorithm in which branching decisions are made with a deep neural network. The proposed exact algorithm utilizes the search order of nodes based on the output of the neural network, with the goal of speeding up the search. Three types of solution representations are compared, along with machine learning models are created for each of them. Computational results confirm the effectiveness of the proposed method, which leads to computation times that are on average around half of those without the neural network. •A B&B with deep neural network is proposed for dynamic berth allocation problems.•Three solution representations are compared along with machine learning models.•The CPU time is reduced on average around half of those without the neural network.
ISSN:0377-2217
DOI:10.1016/j.ejor.2024.06.040