Multi-stage hybrid multiobjective evolutionary algorithm for multi-type vehicle routing problem with simultaneous pickup and delivery and time windows
The vehicle routing problem (VRP) is a fundamental and extensively studied problem in logistics and transportation. Despite its significance, real-world applications often require more complex variants to address specific operational constraints and objectives. To address these limitations, this pap...
Gespeichert in:
| Veröffentlicht in: | Computers & industrial engineering Jg. 207; S. 111242 |
|---|---|
| Hauptverfasser: | , , , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
Elsevier Ltd
01.09.2025
|
| Schlagworte: | |
| ISSN: | 0360-8352 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | The vehicle routing problem (VRP) is a fundamental and extensively studied problem in logistics and transportation. Despite its significance, real-world applications often require more complex variants to address specific operational constraints and objectives. To address these limitations, this paper introduces the multi-type vehicle routing problem with simultaneous pickup and delivery and time windows (MTVRPSPDTW). Although hybrid multiobjective algorithms have been successful in combination optimization problems in recent years, it is still challenging to improve the performance of the algorithms by combining them with different timing. We propose a multi-stage hybrid evolutionary multi-objective optimization with a multi-region sampling strategy (MS-HEMO-MRSS) to optimize both vehicle number and wait time of MTVRPSPDTW. The algorithm integrates a three-stage hybrid approach, combining a global search using the multi-region sampling strategy (MRSS) with a local search based on routing sequence differential evolution (RSDE). The initial stage employs MRSS to quickly position the population near the Pareto front from various directions. The second stage utilizes RSDE to accelerate convergence towards central and edge areas of the Pareto front. In the final stage, individuals on Pareto front are selected and RSDE is used again to guide them towards the edge regions to enhance distribution performance. Specialized encoding, decoding techniques, and genetic operators tailored for two vectors are introduced to optimize MTVRPSPDTW. Comparative experiments with traditional multiobjective evolutionary algorithms demonstrate significant convergence and notable distribution performances, highlighting the benefits of employing different optimization strategies at different evolutionary stages.
•The algorithm integrates a MRSS-based global search with a RSDE local search.•Global search using Pareto dominated relationship-based fitness function and VEGA.•The route sequence differential evolution as local search in different evolutionary stages.•Three stages local search strategies cooperate with multi-region sampling strategy.•Specific coding and decoding methods, tailored to two vectors. |
|---|---|
| ISSN: | 0360-8352 |
| DOI: | 10.1016/j.cie.2025.111242 |