Dynamic-routing in a packet-switched telecommunication network with Quality of Service (QoS) capabilities is addressed. The problem is posed in an informationally decentralized (team) setting, where routing and scheduling decisions are combined. Such decisions are taken at the network nodes, on the basis of local information and possibly of some data received from the neighboring nodes, with the common goal to minimize the expected total delay, spent by packets in traversing the network. Stationarity of the control strategies over an infinite optimization horizon (in the presence of no changes in the traffic parameters and network topology) is achieved by considering an approximation based on a receding-horizon approach. Optimal strategies in this setting are in turn approximated by means of feed-forward neural networks. The problem is posed and a computationally decentralized algorithm for its numerical solution is described. A specific numerical example is also considered, where the neural approximators are tuned, and then used in constructing dynamically varying routing tables and scheduling coefficients in a network simulation based on ns-2, where the gain of the dynamic strategies is evaluated, over an adaptive routing approach based on the measurement of aggregate traffic parameters.

A neural network solution to QoS-IP team-optimal dynamic routing / M. Baglietto; G. Battistelli; R. Bolla; R. Bruschi; F. Davoli; R. Zoppoli. - STAMPA. - (2005), pp. 7452-7459. ((Intervento presentato al convegno 44th IEEE Conference on Decision and Control and European Control Conference 2005 tenutosi a Seville, Spain [10.1109/CDC.2005.1583364].

A neural network solution to QoS-IP team-optimal dynamic routing

BATTISTELLI, GIORGIO;
2005

Abstract

Dynamic-routing in a packet-switched telecommunication network with Quality of Service (QoS) capabilities is addressed. The problem is posed in an informationally decentralized (team) setting, where routing and scheduling decisions are combined. Such decisions are taken at the network nodes, on the basis of local information and possibly of some data received from the neighboring nodes, with the common goal to minimize the expected total delay, spent by packets in traversing the network. Stationarity of the control strategies over an infinite optimization horizon (in the presence of no changes in the traffic parameters and network topology) is achieved by considering an approximation based on a receding-horizon approach. Optimal strategies in this setting are in turn approximated by means of feed-forward neural networks. The problem is posed and a computationally decentralized algorithm for its numerical solution is described. A specific numerical example is also considered, where the neural approximators are tuned, and then used in constructing dynamically varying routing tables and scheduling coefficients in a network simulation based on ns-2, where the gain of the dynamic strategies is evaluated, over an adaptive routing approach based on the measurement of aggregate traffic parameters.
Proceedings 44th IEEE Conference on Decision and Control and European Control Conference 2005
44th IEEE Conference on Decision and Control and European Control Conference 2005
Seville, Spain
M. Baglietto; G. Battistelli; R. Bolla; R. Bruschi; F. Davoli; R. Zoppoli
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2158/348836
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? 3
social impact