|
Load-dependent vehicle routing with route time limit for e-commerce logistics: models and algorithms |
|---|---|
| รหัสดีโอไอ | |
| Title | Load-dependent vehicle routing with route time limit for e-commerce logistics: models and algorithms |
| Creator | Nguyen Thuy Trang |
| Contributor | Pham Duc Tai, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2568 |
| Keyword | Load-dependent vehicle routing problem, Route time limit, Mixed-integer linear programing, Customized heuristic algorithm, Genetic algorithm, Distance-based decoder |
| Abstract | This research explores a class of vehicle routing problem originated from the delivery operations for chilled and frozen food ingredients (e.g., vegetables, meat, and seafood) of an e-commerce platform in Thailand. The objective is to minimize the total load-dependent distance with respect to route time restriction. A mixed-integer linear programming model for the problem is formulated with a base load that represents the weight of an empty vehicle when it is on the final leg of a delivery route. While the model is optimally solved for small-scale instances, it struggles to find solutions for larger instances. To address this challenge, two solution algorithms are developed. The first is a heuristic implementing a customized savings algorithm for route construction in combination with a series of route improvement mechanisms. The second is a genetic algorithm with a distance-based decoder, which dynamically partitions customer se¬quences by load-dependent distances rather than relying on capacity and/or time win¬dow limits. The performance of the proposed mathematical model, and the two algo¬rithms are evaluated using benchmark instances that are adapted from standard data repositories. These modified datasets feature different vehicle capacities, route time limits, and base load levels. Computational results demonstrate that the two proposed algorithms achieve near-optimal performance with the genetic algorithm showing competitive performance and often giving better solutions than heuristics, especially in small and moderate problem instances. In addition, the results show that the base load level influences the objective function and the performance of both algorithms. In particular, moderate base load level results in the best performance of the algorithms. |
