진학프로석·박사 채용 접수 플랫폼

Founded in 1843, IMT Mines Alès currently has 1,400 students (including 250 foreign students) and 380 staff. The school has 3 research and teaching centres of a high scientific and technological level, working in the fields of materials and civil engineering (C2MA), the environment and risks (CREER), artificial intelligence and industrial and digital engineering (CERIS). It has 12 technology platforms and 1,600 partner companies.

The person recruited will be involved in the Trusted Autonomous and Self-Adaptive Systems for Defence (SAACD) project. This project aims to overhaul the engineering and development of this type of complex system.

A SAACD can be defined at two levels:

• • SAACD Component: This is a UAV made up of hardware and software sub-systems, capable of observing, predicting, deciding and reconfiguring itself to fulfil its mission (e.g. surveillance, detection, tracking, control, etc.). The integration of on-board AI requires new engineering methods, which are still poorly supported by digital twins.
• • SAACD System of Systems (SdS): this corresponds to the assembly of several SAACD Components, in homogeneous swarms or heterogeneous packs to fulfil a common mission, e.g. monitoring a critical infrastructure.

Both must be designed to promote these autonomy and self-adaptation capabilities. Three major challenges have been identified:

• • (P1) modelling uncertain environments where robust, weakly supervised machine learning algorithms can be deployed to irrigate engineering processes and validate this design as early as possible;
• • (P2) integrate a reliable and frugal embedded AI (limited to inference and/or also capable of learning, e.g. reinforcement learning) as one of the components of a SAACD. A component that facilitates, again with a demonstrable level of confidence, observation, prediction and the decision to adapt or partially reconfigure;
• • (P3) Ultimately, aim for SAACD certification based on proof of validity and reproducibility.

The R&D questions concern :

• • (For P1) "AI as a service" to enrich engineering: modelling of the changing, complex and uncertain environment (essentially for the purposes of assisting understanding, verification and validation) of the uncertain environment in which the SAACD Component or the SAACD SoS evolves;
• • (For P2) "AI as a component" as a component in its own right of the SAACD, but which must reconcile frugality and reliability of the actions envisaged according to the situation encountered in this moving and unpredictable environment;
• • (For P3) Exploit the Digital Twin Systems, integrating the AI contributions developed in P1 and P2, to optimise the simulations, certification, maintenance, training and piloting of the two types of SAACD.

This project will involve formalising and integrating the contributions of : (i) model-based and data-driven systems engineering (MBSE), (ii) embedded frugal (explainable and robust) AI, and finally (iii) advanced modelling, simulation and optimisation techniques in a complex uncertain environment for the design of a SAACD.

The working approach is based on four successive and iterative phases, making it possible to demonstrate the feasibility of the SAACD project. The person recruited will therefore be involved in each phase, which combines methodological, technological and experimental aspects, drawing on the expertise of SyCoIA's Lecturers/Researchers in MBSE/MBSSE, predictive analytics, autonomous decision-making, robust and explainable AI, and interoperable digital twins. These phases are summarised below:

• • Phase 1 - Definition and modelling of use cases: M1-M3
• • Phase 2 - AI services for engineering (AI-as-a-Service): M4-M8
• • Phase 3 - Trusted frugal embedded AI (AI-as-a-Component) : M6-M11
• • Phase 4 - Integration, demonstration and proof by digital twins: M10-M16