Ph.Ds Hiring

This Ph.D. will be supervised by Sophie Ebersold and Jean-Michel Bruel.

This Ph.D. will be supervised by Marc Pantel, Sophie Ebersold and Jean-Michel Bruel.

Nowadays, systems are becoming increasingly complex. This is due to several factors:the integration of complex components into systems, the consideration of systems of systems, the heterogeneity both in the systems (a system is the composition of several software, hardware and interface parts with the external world) and in their environment (for example, an avionics system will operate according to the laws of aerodynamics, the rules of air navigation and safety and security and many other constraints), the integration of technologies derived from Artificial Intelligence (AI), the involvement of several stakeholders in the development of systems, etc. In an extended team development framework, such as that of the Extended Enterprise [13], a major work requiring the involvement of several stakeholders introduces an additional complexity of communication and collaboration between the different stakeholders. This increasing complexity requires a shift towards digitization of development processes. The goal is to better manage the different tasks as well as the documents and models handled and/or produced throughout the development process. Digitization consists in moving from the "paper" and semi-formal version of documents to a formalization and automation of these different elements (documents and manual steps). Among the objectives of this digitization are the acceleration of the development time and the assurance of the quality of the system under development. Continuous Integration and Continuous Deployment (CI/CD) [15, 16] are key activities in a development with DevOps practices[18]. Model-Based Systems Engineering (MBSE) aims to shift traditional document-centric development work to model-centric development. From a MBSE development perspective, the deployment of the DevOps and CI/CD approaches helps to better manage the complexity of systems.

The correctness of a system is ensured through verification and validation (V&V) activities. Wasson [1] defines: (i) the verification as “the continuous process of evaluating the products of the system acquirer, system developer, service provider, or user throughout the system/product life cycle to assess compliance with predefined contract, mission, or task requirements”, (ii) the validation as “the continuous process of evaluating and analyzing the extent to which multidisciplinary Systems Engineering (SE) activities and work products -decision artifacts such as specifications, designs, or devices-satisfy the user based on the user’s predefined business needs, constraints, and performance expectations. In other words, verification answers the question "are we developing the system right?" (i.e., in compliance with the appropriate rules or quality requirements) and validation aims to answer the question "are we developing the right system?" (the one that the customer really needs and that complies with his requirements). Correctness in relation to the specifications is one of the primary criteria defining the quality of the system. Moreover, the earlier this criterion is taken into account in the development process, the less expensive it will be to deal with errors. A late correction of errors can cost up to 200 times more than a correction made earlier in the requirements analysis phase [2]. V&V is a rigorous process that must be done upstream and as development of the system. It starts from the elicitation and analysis of the requirements until the final system is obtained. The importance of these activities/processes becomes apparent when it comes to the development of critical and complex systems such as those in the transportation field (airplanes, autonomous cars, etc.) or the medical field (pacemakers, insulin pumps, hemodialysis). The objective of development based on a DevOps approach is to produce and integrate models in a continuous manner. The cost of verifying some properties inside models or between models during this continuous integration in DevOps is far from insignificant. Such a continuous evaluation is very costly, especially when moving from the scale of software to complex systems containing heterogeneous parts. This V&V cost does not allow to replay all activities at any time, unlike testing, typing, etc. The question is to identify the technological means of implementation for scaling: should an OCL interpreter be used? Or would it be more appropriate to implement it in a more efficient language? And what about automatic translation to a more suitable language or toolchain? etc.

The International Council On Systems Engineering (INCOSE) defines model-based systems engineering (MBSE) as "the formalized application of modeling to support system requirements system requirements, design, analysis, verification and validation activities beginning in the validation activities beginning in the design phase and continuing throughout the development and development and later phases of the life cycle." (INCOSE SE Vision 2020). This work is accomplished by following a methodology, adopting one or more tools and applying one or more This work is accomplished by following a methodology, adopting one or more tools and applying one or more modeling languages such as SysML® [7].