This study addresses a real-world agricultural logistics problem arising from an industrial request: the optimization of the collection and delivery of containers filled with agricultural products (e.g., tomatoes) distributed across known geographic locations to processing facilities, using a fleet of trucks with limited capacity, deployed across multiple depots. The objective is to minimize the total distance traveled by the fleet while ensuring that all containers in the area are collected. The problem is modeled as a variant of the well-known Capacitated Vehicle Routing Problem (CVRP), extended to include multiple pickups, initial depots, final delivery points, and binding load-capacity constraints. Three solution approaches were developed: a greedy heuristic algorithm based on field proximity, and two metaheuristic algorithms based on Ant Colony Optimization (ACO). In the first ACO, routes are influenced by pheromone trails while the collected quantity remains greedy; in the second, the number of containers collected at each location is also optimized. The algorithms were first validated on a toy model to verify correctness and stability. Subsequently, systematic experiments were conducted to calibrate the number of iterations and the number of fleets to simulate, and to determine the optimal hyperparameter configuration for the ACO algorithms. Finally, the three approaches were compared in simulated scenarios and on real-world data provided by the industrial partner. Datasets (including anonymized real-world data) and experimental scripts are made available in open-source format to ensure research reproducibility. The results demonstrate that both ACO-based metaheuristics consistently outperform the greedy heuristic in terms of total distance traveled. While the first ACO achieves superior performance in approximately 50% of the cases and the second in around 20%, with 31% of best solutions identified by both, the first generally ensures more consistent convergence to high-quality outcomes. However, the advanced ACO, by dynamically optimizing the collected quantity, demonstrates decisive advantages in specific scenario configurations, highlighting its role as a flexible complement to the basic approach. These findings highlight the importance of balancing route optimization with adaptive collection decisions and confirm the practical applicability of the proposed methods in real-world agricultural logistics scenarios.
Multi-depot Capacitated Vehicle Routing Problem with Processing Facilities (MDCVRP-PF): a comparative study of heuristic and Ant Colony Optimization-based metaheuristics for agricultural logistics
Sgarro, Giacinto Angelo
;Colasanto, Francesco;Grilli, Luca
2026-01-01
Abstract
This study addresses a real-world agricultural logistics problem arising from an industrial request: the optimization of the collection and delivery of containers filled with agricultural products (e.g., tomatoes) distributed across known geographic locations to processing facilities, using a fleet of trucks with limited capacity, deployed across multiple depots. The objective is to minimize the total distance traveled by the fleet while ensuring that all containers in the area are collected. The problem is modeled as a variant of the well-known Capacitated Vehicle Routing Problem (CVRP), extended to include multiple pickups, initial depots, final delivery points, and binding load-capacity constraints. Three solution approaches were developed: a greedy heuristic algorithm based on field proximity, and two metaheuristic algorithms based on Ant Colony Optimization (ACO). In the first ACO, routes are influenced by pheromone trails while the collected quantity remains greedy; in the second, the number of containers collected at each location is also optimized. The algorithms were first validated on a toy model to verify correctness and stability. Subsequently, systematic experiments were conducted to calibrate the number of iterations and the number of fleets to simulate, and to determine the optimal hyperparameter configuration for the ACO algorithms. Finally, the three approaches were compared in simulated scenarios and on real-world data provided by the industrial partner. Datasets (including anonymized real-world data) and experimental scripts are made available in open-source format to ensure research reproducibility. The results demonstrate that both ACO-based metaheuristics consistently outperform the greedy heuristic in terms of total distance traveled. While the first ACO achieves superior performance in approximately 50% of the cases and the second in around 20%, with 31% of best solutions identified by both, the first generally ensures more consistent convergence to high-quality outcomes. However, the advanced ACO, by dynamically optimizing the collected quantity, demonstrates decisive advantages in specific scenario configurations, highlighting its role as a flexible complement to the basic approach. These findings highlight the importance of balancing route optimization with adaptive collection decisions and confirm the practical applicability of the proposed methods in real-world agricultural logistics scenarios.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
