The goal of routing in a telecommunications network is to distribute user traffic from in a network of sources and destinations in accordance with the traffic's service requirements and subject to the network's resource restriction. Objectives of routing scheme include maximizing the network performance, such as larger carried load, fewer blockings and fewer crankbacks, while minimizing the cost, such as fewer trunk groups and fewer intermediate tandems. Constraints are imposed by the limitations of trunk resources, user traffic, and the network services provided and user services requested....[ Read more ]

The goal of routing in a telecommunications network is to distribute user traffic from in a network of sources and destinations in accordance with the traffic's service requirements and subject to the network's resource restriction. Objectives of routing scheme include maximizing the network performance, such as larger carried load, fewer blockings and fewer crankbacks, while minimizing the cost, such as fewer trunk groups and fewer intermediate tandems. Constraints are imposed by the limitations of trunk resources, user traffic, and the network services provided and user services requested.

Mean field and Monte Carlo computer simlutaions programmes are built according to the queueing theory and telecommunication engineering theory. Several traffic processes, such as blockings, crankback and overflow, and a variety of traffic profiles, topologies and structures are under investigation. We also develop a new software package TNSP (Telecom Network Simulation Package) to simulate various traffic conditions of East Kowloon telephone network of Hong Kong.

As the network technologies evolve, enhanced routing methods are required to meet the dramatic demand of the growth of telecommunications services. We in-vent a new adaptive and efficient routing technique called Neural Routing to tackle this stochastic and nonlinear optimization problem. We use a simplex-based central-ized algorithm to dynamically distribute telephonic traffic among alternate routes in circuit-switched networks according to the fluctuating number of free circuits and the evolving call attempts. It gives an excellent network performance especially in asymmetric traffic conditions and uneven network configurations. We apply it to gen-erate examples for training localized Neural controllers. Simulations show that the decentralized Neural approach has a significant improvement in blocking probability with Maximum Free Circuit (MFC) which forms the basis of AT&T's Real-Time Net-work Routing (RTNR) and gives a surpassing performance in crankback percentage especially for an unbalanced configuration network with uneven traffic profile.