Présentation sur YouTube.

The ROADEF challenge, hosted by the French Operations Research and Decision Support Society (ROADEF) and EURO, is a bi-annual competition for researchers and practitioners in the field of optimization. The 2022 challenge features a large-scale 3D truck loading problem in collaboration with industrial partner Renault Group. Here, the goal is to combine a set of items into stacks and load these into trucks, as to minimize total transportation and inventory costs. We have developed a scalable column generation heuristic for this problem, in which the pricing problems correspond to single truck loading problems. We heuristically solve these problems using a labelling algorithm on a dynamically defined stack graph. We warm-start the algorithm using the solution from a simple assignment problem, which also provides lower bounds on the optimal solution, and perform several post-optimization steps. In this talk we present our methodology, the results of the final round of the challenge and a discussion of the limitations. This version of the method received a first-place award in the junior category, dedicated to researchers who have not yet completed their PhD.