Real-time decentralized traffic management using a parallel algorithm

A decentralized, state-dependent access-control and routing strategy for circuit-switched networks is presented. To decentralize the control, the network is partitioned into subnets and a specific traffic controller is assigned to each one. The controllers obtain periodic subnet measurements, and compute the optimal control policy through an iterative and parallel dialog with other controllers. The following objectives for traffic control are considered. First, to allocate the incoming demand, maximizing the predicted minimum trunk group residual capacity over the network. Secondly, when the projected demand cannot be accommodated, optimal strategy should reject fairly the extra demand at source. The mathematical formulation of the above objectives leads to an equilibrium programming problem (EPP). It is shown that the EPP can be decomposed into a number of subproblems solved in parallel by the intercommunicating controllers