The verification of complex engineering systems from the very early phases of the design process is of primary importance, as it directly influences performance and system functionalities. Traditional design approaches aim at using simulations as a set of tools during the verification process. However, the current trend in the industry is towards simulation-based design processes in an iterative manner so as to constantly evaluate the system development. This perspective conveys the design process towards a verification-based design process. In the very early phases of the design process, evaluating different concepts for further development is not without problems, since a certain amount of product information is missing in the early phases. Therefore, traditional approaches have aimed at considering expert's opinions as the main evaluation criteria for assessing pre-concepts and concept designs. However, qualitative-based methods are highly limited according to expert's subjective judgements, level of expertise, as well as the ability to take into account multidisciplinary criteria in the case of complex systems.<br /><br />This dissertation presents research work related to the verification-driven design process of complex mechatronic systems using a stochastic reliability method for evaluating the concept design from the early phases of the product development. The main objective of this thesis consists in demonstrating the advantages of an innovative system design process based on a quantitative evaluation method using reliability as the main criteria. This thesis reviews the state of the art of the verification and validation process, describes different trends in the system design processes towards simulation-based design processes and reviews the best practices of decision-making processes in the engineering field. The work conducted during this thesis consists of the development and modelling of the verification-driven design approach. The method uses the stochastic Petri Net approach for modelling the operational and functional sequence of the system as well as its dysfunctional behaviour. Reliability parameters of each concept are estimated based on their level of design and thus various concepts can be evaluated against each other.<br /><br />The method is applied to case studies that consist of the development of a Remote Handling system for the maintenance of a fusion reactor called DEMO. The results confirm the benefit of such a method for designing and evaluating the concept design from the very early phases of the system development. The purpose of this research is to maintain the usefulness of the findings for other developments at a larger scale and in other fields than fusion engineering.