Optimizing Multi-Depot Vehicle Routing Under Dynamic Demand Conditions Using a Learnheuristic Approach

In industries where production centers depend on a steady supply of materials, planning logistics can become complex due to shifting demand patterns. Factors like seasonal changes, regional events, and varying consumer trends contribute to dynamic demand, making it challenging to maintain efficient routes for decision-makers. This paper addresses the Multi-Depot Vehicle Routing Problem (MDVRP) under these dynamic demand conditions, and it presents a learnheuristic methodology that combines a metaheuristic algorithm with a supervised machine learning model. The machine learning model estimates the demand fluctuations and provides this information into the solution construction process rather than relying solely on static demands. The computational experiments demonstrate the learnheuristic approach provides near-optimal solutions that account for demand fluctuations, optimizing routes across multiple depots and minimizing stockouts. This offers an interesting alternative for decision-makers in practical applications to streamline logistics planning and reduce operational costs.