Team theory and neural approximators for dynamic routing in communication networks

Abstract

The dynamic-routing problem in communication networks is addressed. The nodes of the network must accomplish the following tasks: 1) generating routing decisions to minimize the expected total delay, spent by messages in the queues at the nodes and on the network links, on the basis of local information and possibly of some data received from other nodes, typically the neighboring ones, and 2) computing (or adapting) their routing strategies by measuring local variables and exchanging a small amount of data with other nodes. The first task regards the nodes as the cooperating decision makers of a team organization. The second task calls for a computationally distributed algorithm. Such tasks and the well known impossibility of solving team optimal control problems under general conditions suggest that each decision maker acting at a node be assigned an axed-structure routing strategy, in which some parameters have to be optimized. Feedforward neural networks have been chosen for their powerful approximation capabilities. Simulations performed on complex communication networks show the effectiveness of the proposed method.