The empty railcars repositioning strategy generates no income but is crucial for a good service quality, it should then satisfy two main objectives: fullling the customer demand and generating as little expense as possible. Moreover, because of breakdown or heavy traffic, variation on travel times happens to be the main cause of uncertainty in railroad scheduling and must be taken into account to suggest a robust planning. This thesis presents the linear program used in a prototype tool for the optimization of empty railcar repositioning strategy designed for the SNCF¹. The resulting schedule is computed with CPLEX and minimizes moving cost, delay and unfulllment penalties. Substitutions of railcar categories are also permitted and eventually penalized. In addition, uncertainty on travel times is handled by considering the expected cost of a move (regarding delay probability and possible penalties) and by adding slack periods at the end of moves. The robustness can be modulated through the use of a cursor. Finally, the model enforces a decision making process previously dened by the SNCF to ensure that the suggested planning can be easily grasped and trusted by users. Schedules have then been generated based on randomly generated data and simulated. Results show a potential saves of 10% on considered costs and a good range of use of the robustness cursor is suggested. Finally, paths for improvement of this prototype are proposed to meet the eventual schedulers' further needs in order to move forward the production of this tool at the company scale. ¹Société Nationale des Chemins de fer Français (French National Railways) / Master of Science