S E M I N A R
Centra za atomsku i subatomsku fiziku
Laboratorija za fiziku atomskih sudarnih procesa
"Fault tolerant control for multi-sensor switching schemes: Invariant set constructions and implications"
Dr. Sorin Olaru
Supélec (École Supérieure d`Électricité) France
utorak, 28.10.2008. u 12 h, velika sala "Zvonko Marić" Instituta za fiziku, Pregrevica 118, Zemun
Saradnja iz programa "Pavle Savic"
Sorin Olaru received his B.S. (2001) and M.S. (2002) in electrical engineering from the University "Politehnica" Bucharest, and the PhD. degree (2005) in Automatic Control from University Paris XI, France. He held a research position in INRIA Rennes and visiting appointments at NTNU Trondheim (Norway) and CSDC Newcastle (Australia). He is currently an assistant professor in Supélec (École Supérieure d`Électricité) France.
Sorin Olaru is the receipt of the EU Archimedes prize 2002 for his work on the predictive control for hybrid systems with applications in irrigations. His research interests include the optimisation-based control techniques, viability and geometrical characterisation of constrained control. Lately he is involved in research projects related to network control systems with results on fault tolerant control and control of systems affected by delays.
In this talk we will present several results on the control on multi-sensor schemes. The control design is based on a switching strategy (among the state estimations) and handles the fault appearance by geometrical separation of the sets characterising the healthy and faulty system functioning.
Positive invariance is a common analysis and control design tool for systems affected by constraints and disturbances. In the present talk, the construction of epsilon-approximations of minimal robust positive invariant sets for linear systems will be revisited. The results are used in the context of multi-sensor control schemes which have to deal with attraction basins and invariant sets for different operating regimes. Indeed, these are specific problems originated by the switching between different sensor-estimator pairs and by the presence
of faults in some of the sensors. Within the same framework, global stability of the switching strategies can be assured if the invariant sets topology allows the exclusive selection of estimates obtained from healthy sensors.
The problem of sensor recovery will be evoked and extensions of the method to nonlinear models are pointed out. Finally, is shown that receding horizon optimisation strategies can be further used to improve the performances and to allow an increased range of fault tolerant functioning.