Models and methods for the optimal distribution in logistics supply chains: application to the Decathlon company

In the past decades, the vehicle routing problem has been a topic of interest to scientists, industrialists, industry owners, governments, and environmentalists, as new measures, policies, and technologies are being developed to reduce greenhouse gas emissions and, more broadly, air pollutants. This PhD, which is a joint project of the University of Genova and Decathlon Italy, aims to optimize the distribution between the final warehouses, called distribution centers, and the stores in the last two levels of the supply chain. To achieve this optimization, several objectives are considered, such as decreasing total distribution costs, reducing air pollution, and at the same time satisfying all demand without delays while considering all constraints of the supply chain, such as store delivery time windows and available trucks with different capacities. To address this type of project, we used mathematical optimization, specifically referring to the capacitated vehicle routing problem. A set of mathematical models has been developed, each with distinct details and constraints to account for various problem scenarios. The first mixed-integer model aims to minimize total cost and emissions as a single objective, in which the costs are derived from a multiple regression model. The second version aims to minimize the total cost and emissions trade cost, in which a mixed fleet (internal combustion engine, electric and hydrogen vehicles) has been considered. The mixed fleet is applied in the mathematical model based on its performance in terms of energy consumption and the resulting costs and emissions. In this second version, emissions are expressed as emission trade costs using the carbon cap-and-trade mechanism. The third version of the model consists of a bi-objective optimization, in which total costs and total emissions are considered as distinct objectives. In the third mathematical model, we consider a longer decision-making horizon, requiring future store demand. To estimate long-term demand, time-series and machine-learning forecasting methods are implemented and analysed. All models are piloted and tested on real cases from Decathlon Italy and France, and their results are compared with current procedures. Finally, analyses and managerial insights are provided to facilitate informed decision-making.