Modellabstraktion

Bestimmung vereinfachter (abstrakter) dynamischer Modelle aus gegebenen komplexen Prozessmodellen. Bei dieser Fragestellung ist von zentraler Bedeutung, dass ein auf der Grundlage des vereinfachten Modells entworfener Regler die Einhaltung der vorgegebenen Anforderungen auch das fuer das urspruengliche komplexe Modell garantiert.

Model abstraction for control purposes

There exists a natural tradeoff between model accuracy and model simplicity. As a consequence, we want a process model to be as precise (and therefore as complex) as necessary to solve a given application problem, but also as simple (and therefore as inaccurate) as possible without sacrificing solvability. This minimizes the workload in any subsequent (model-based) synthesis or design step. In feedback control applications, there is a particularly large scope for model approximation, or abstraction, as feedback constitutes an extremely efficient way of "combatting" model uncertainty. A central requirement for model abstraction with regard to control applications is "safety": a safe approximation is characterized by the fact that each controller enforcing the specifications for the simplified model will also ``work properly` for the underlying system. This represents joint work with Siu O`Young, Memorial University, St.John`s, Canada and Thomas Moor, Australian National University, Canberra.