| Lecturer Profile |
René K. Boel received a degree in electromechanical engineering (1969) and in nuculear engineering (1970) from Ghent University, a M.Sc. (1972) and a Ph.D. degree (1974) in Electrical Engineering and Computer Sciences from the University of California, Berkeley. He is currently a professor at the School of Engineering, Ghent University, Ghent (Belgium). He was a research fellow of the F.W.O.-Vlaanderen (Flemish Foundation for Scientific research) until September 2000. He has held temporary appointments at the Department of Electrical Engineering and Computer Sciences, University of California Berkeley, at the department of technical Cybernetics NTH-Trondheim Norway, at the Mathematics department of the Katholieke Universiteit Leuven, at the Mathematics and Statistics department of Bell laboratories Murray Hill (N.J.), at the department of Computing and Control of Imperial College London, at the Department of Systems Engineering, Australian National University, Canberra, at the department of Electrical and Computer Engineering, University of Newcastle, Australia, and at the Electrical and Electronic Engineering department, Melbourne University.
His current research interests are in the fields of distributed and supervisory control and in estimation and fault detection for stochastic, discrete event and hybrid systems, with applications to traffic networks and power systems, as well as in bio-informatics. |
| Lecture Abstract |
Optimal feedback control usually tries to minimize the deviation of the state of the plant from its desired reference trajectory, taking into account the cost of the control actions. The performance of a closed loop control system depends on the accuracy of the state estimates available for the feedback control action. In these talks we will consider how the cost of activating different sensors can be included into this control decision problem. Activating sensors can be costly due to power limitations (e.g. battery powered sensors) or due to communication constraints (remote sensors connected to the controller via limited capacity communication links). The cost of activating sensors must be compared to the performance improvement the additional information achieves in order to obtain an optimal design.
The talk will present a number of practical cases where the issue of sensor activation is important. It will be shown that in the case of linear quadratic Gaussian systems the optimal control design problem can be decomposed in 2 simpler subproblems, and that the feedback control decision for the actuator remains the same as if no sensor activation control is performed. For more general systems a heuristic will be presented that achieves a control performance with a known upper bound. The emphasis in this talk is on random, stochastic systems, but discrete event models will also be considered.
Examples are taken from manufacturing systems, traffic control, smart power grids, and medical applications. |
