Abstract
Pickup and delivery problem (PDP) and dynamic vehicle routing problem (DVRP) are two key components of crowdsourced freight delivery services. Although previous research has focused predominantly on static vehicle routing problems, this study formally defines the dynamic problem specific to crowdsourced freight delivery and presents a mixed-integer linear programming model based on a rolling-horizon framework. The objective is to minimize total service costs, including fixed vehicle costs, transportation costs, and penalty costs for delays, while planning routes that cover all orders. To solve this combinatorial optimization problem, we propose an improved partheno genetic algorithm (IPGA) and a simulated annealing algorithm (SA). Numerical experiments demonstrate that the IPGA outperforms the SA, reducing the total service costs by over 10% on average. In addition, a real-world case study illustrates the practical applicability of our model and algorithms, providing a solid foundation for real-world implementation.