Ankündigung eines Vortrags im Graduiertenkolleg


"From Discrete Event Systems to Continuous models:
A Petri Nets based coherent view"

3. Dezember 2001

Prof. Manuel Silva
Departamento de Informatica e Ingeniería de Sistemas
Centro Politécnico Superior, Universidad de Zaragoza

Montag, 3. Dezember 2001, 10 Uhr
TU Berlin, Einsteinufer 17, Gebäude EN, Raum 719
Fluidification of Petri net models leads to continuous and hybrid systems. Surprisingly, not all DEDS models can be fluidified. In fact, fluidification requires "big populations" and DEDS are not always behaviourally monotonic with respect to population homothetic growth.

First-order fluidification of stochastic (discrete) PN models leads to deterministic and continuous models. Nevertheless, basic qualitative properties, like absence of deadlocks are not preseved in general (e.g. you may have a live and bounded discrete time-stochastic model, while the corresponding continuous model is unbounded and non-live!). Therefore results concerning continuisable net subclasses are required.

Another interesting reflection concerning continuous net models is that, even if the continuous approximation (or view) is stable, it may happen that no steady-state exists (a clearly non-linear behaviour for the continuous relaxed model). This is another "unexpected" property, in the sense that works in the field of continuous models usually assume its existence.

The presentation will also introduce some design problems, like the computation of a minimal initial marking, that surprisingly happens to be polynomial for certain net subclasses. Some control results will be introduced, making a bridge to scheduling in the underlying discrete model.