Tid: 14 maj 1998 kl 1315-1415 (Observera dagen och tiden!)
Plats : Seminarierummet 3721 (Obs platsen!), Institutionen för matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Evsei Morozov, Dept. of Applied Mathematics & Cybernetics, Petrozavodsk State University. (Publikationslista)
Titel: Weak regenerative queueing network processes
Sammanfattning:
It is impossible usually to study complex queueing networks by pure analitical technique (only exception is a narrow class of the so-called Markovian networks, in which case exponential distributions are used). Moreover, there are hard problems when we attempt to apply classic statistical methods to analyse queueing network processes by simulation because of initial (transient) period and strong correlations between adjacent values of the process.
In this connection the regenerative approach turns out often to be very useful to supply sufficiently correct analysis of the network process.
We will discuss both classic and the so-called weak (wide sense) regeneration,where a dependence between adjacent regeneration cycles is allowed. We also give some typical simple examples of regenerative queueing processes (including weak ones) and dicsuss the assumptions which guarantee the existence of the regenerative structure.
We suppose also to show how to construct the confidence interval to estimate the limit characteristics of the process (also including weak regeneration case). Finally, we discuss possible advantages of weak regeneration in comparison with classic regeneration.
The renewal technique and the (extended) CLT are the main tools of our analysis.
Key words: queueing network, regenerative process, embedded renewal process, confidence interval, Central Limit Theorem for m-dependent variables