Solution algorithms for dock scheduling and truck sequencing in cross-docks: A neural branch-and-price and a metaheuristic
In this study, we present a novel mathematical model for the integrated challenge of dock scheduling and truck sequencing in cross-docking facilities. Many existing models for variations of this problem in the literature rely on big-M constraints, known for their poor performance when applied to gen...
Uložené v:
| Vydané v: | Computers & operations research Ročník 167; s. 106604 |
|---|---|
| Hlavní autori: | , , |
| Médium: | Journal Article |
| Jazyk: | English |
| Vydavateľské údaje: |
Elsevier Ltd
01.07.2024
Elsevier |
| Predmet: | |
| ISSN: | 0305-0548, 1873-765X |
| On-line prístup: | Získať plný text |
| Tagy: |
Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
|
| Shrnutí: | In this study, we present a novel mathematical model for the integrated challenge of dock scheduling and truck sequencing in cross-docking facilities. Many existing models for variations of this problem in the literature rely on big-M constraints, known for their poor performance when applied to general-purpose mixed-integer programming solvers. This introduces significant limitations, especially when tackling smaller instances. Consequently, most solution methods in the literature gravitate towards metaheuristic approaches. Our proposed model offers a compact formulation without any big-M constraints, utilizing 4-index variables. It is strategically designed to lend itself well to a dual decomposition approach (Dantzig–Wolfe) and can be effectively solved using a branch-and-price methodology. We address the pricing problem through a branch-and-cut scheme and illustrate its efficiency in handling large instances. Recognizing the pricing problem as the primary bottleneck, we employ a deep neural network trained on a comprehensive set of instances with the same distribution to predict promising duals. Extensive computational experiments demonstrate a notable reduction of over 50% in overall computational times. Additionally, we introduce a heuristic sharing similarities with certain Variable Neighborhood Search (VNS) approaches. Our proposed heuristic has proven highly efficient in extensive computational experiments.
•A new IP model for the cited problem in cross docking is presented with O(n4) variables.•An efficient branch-and-price as well as a VNS method are proposed to solve the problem.•A deep neural network is trained to predict promising duals, in the process of Branch-and-Price.•Computational experiments shown over 50% reduction in the overall computational times.•This study is showcasing the synergy between operations research and machine learning. |
|---|---|
| ISSN: | 0305-0548 1873-765X |
| DOI: | 10.1016/j.cor.2024.106604 |