An Adaptable Heuristic Framework for Multiproduct Vehicle Routing Problem with Intra-Route Constraints

This thesis presents mathematical framework for solving transport planning problems in manufacturing supply chain. The work is motivated by a real-world use case involving the collection of goods from a network of suppliers to a central production facility. Rather than focusing only on optimizing a specific scenario, the goal is to develop a generalizable solution method capable of adapting to a variety of similar logistics contexts. To this end, the thesis introduces both an exact optimization model and a heuristic algorithm. The exact model is based on a variant of the Capacitated Vehicle Routing Problem (CVRP), namely the Split-Delivery Vehicle Routing Problem (SDVRP) and is implemented using Python and Google OR- Tools on the Azure Databricks platform. The heuristic method, designed for scalability and flexibility, employs rule-based modules (also called controls) to create both feasible loads and routes. This modular approach allows each control to be adaptable to the specific requirements of each use case. The proposed framework is tested on a range of synthetic instances featuring uniformly distributed suppliers and varying problem sizes, with the aim of evaluating its performance in a relatively generic scenario. It is compared to a decomposition heuristic that separatesloading and routing into two distinct phases. The results reveal that, under the tested settings, the two heuristic approaches lead to comparable performance. Future work could explore the adaptability of the proposed method to different logistical environments, such as those involving clustered supplier locations, and different probability distributions for the features of the products.