|
 |
|
 |
Modelling, Optimisation and Control of Systems |
 |
The « Modelling, Optimisation and Control of Systems» (MOCOSY) scientific area federates the activities of four research groups: Diagnosis and Supervisory control (DISCO), Methods and Algorithms in Control (MAC), Modelling, Optimisation and Integrated Management of Systems of Activities (MOGISA), Modelling and Control of Networks and Signals (MRS).
The Mocosy scientific area deals with uncertain dynamic systems within the framework of control and production systems. The research goals summarize in the observation, the control and the optimisation of such systems, approached at different levels of abstraction and along different architectures. The main key-words are then estimation, surveillance, diagnosis, state tracking, control, planning, scheduling, routing and, as a common factor, modelling. Strong links exist with the Applied Maths and the Artificial Intelligence (AI) fields.
The originality of our research stands on two features: it is guided by flexibility and robustness needs and it is positioned at the crossing of several scientific domains. The following topics come out from our expertise scope:
• Provide solutions to scheduling, planning and resource management and scaling problems: original research is performed on global and discrete optimisation, combining combinatorial optimisation (Operations Research domain) and constraint programming (AI domain), as well as on the underlying algorithmic issues.
• Moments, optimisation and algebraic geometry: semi-definite programming is one of our tools and drives us towards optimisation techniques, in particular non-convex optimisation, supported by the theoretical foundations of polynomial moments.
• Diagnosis theories: the originality of our research lies on making profit of diagnosis approaches proposed in the control field and on logical model based diagnosis theories in the field of AI. Several significant results can be outlined for the monitoring and diagnosis of continuous systems, discrete event systems, and hybrid systems. A particular interest is given to the properties for the design of diagnosable and self-healing systems.
• Filtering and detection theories: the outstanding contributions along this topic are our original results on the random and the deterministic approaches to particle filtering, as well as work in Volterra non-linear filtering.
• System identification and realisation theories: an original method has been developed to obtain the exact realisation for linear and bilinear dynamic systems. This technique, inspired from optimal filtering theory, lies on building a hereditary output predictor, i.e. whose memory increases with data grow. This method has been successfully extended to the identification problem.
• Control theories: our contributions focus on linear and non-linear control, with emphasis on constructive methods and synthesis of structured control laws in state space or frequential frameworks. Original results can be outlined for the multi-objective linear structured control problem, the windup problem, the design of control laws for periodic or hybrid systems, PDEs.
• Diffusive representation and non linear operators: this theory is devoted to analysis, approximation and synthesis of a wide class of dynamic operators, providing the required extension for dealing with phenomena involving a high number of dynamic variables, for example distributed systems. Powerful state space formulations follow from a dual representation, which are well-adapted to practical problems thanks to their numerous properties inherited from underlying diffusive equations.
|
|
|
|
 |
|
