Abstract
We consider a robust inventory problem where products are perishable with a given shelf life and demands are assumed uncertain and can take any value in a given polytope. Interestingly, considering uncertain demands leads to part of the production being spoiled, a phenomenon that does not appear in the deterministic context. Based on a deterministic model we propose a robust model where the production decisions are first-stage variables and the inventory levels and the spoiled production are recourse variables that can be adjusted to the demand scenario following a FIFO policy. To handle the non-anticipativity constraints related to the FIFO policy, we propose a non-linear reformulation for the robust problem, which is then linearized using classical techniques. We propose a row-and-column generation algorithm to solve the reformulated model to optimality using a decomposition algorithm. Computational tests show that the decomposition approach can solve a set of instances representing different practical situations within reasonable amount of time. Moreover, the robust solutions obtained ensure low losses of production when the worst-case scenarios are materialized.