An obstacle to the adoption of model-checking in large projects is a lack of guidelines on how to integrate formal methods with existing system engineering practices. In this context, a methodology should give answers to several questions: How to manage the models and abstractions used to verify a claim? How do we gain confidence on the soundness of these models? How can we build a structured argument from the verification results? In this paper , we describe a structured approach for managing verification arguments an apply it to check a critical function of an autonomous rover.