Punctuality and reliability are essential elements of passenger transportation system. But disruptions such as infrastructure failures and rolling-stock breakdowns can occur, and then require an adaptation of the timetable, rolling-stock and crew schedules. The aim of this study is to anticipate those potential disruptions by proposing robust rolling-stock rosters. A robust rolling-stock roster should anticipate operational disruptions in order to limit service quality deterioration and additional costs. So we introduce a criterion to measure robustness of a rolling-stock roster. We model the robust rolling-stock planning problem by a mixed-integer linear program (MILP). This problem is di cult to solve because of its exponential number of variables. Then, we obtain a good feasible solution by a column generation-based approach.