Predictive congestion control of ATM networks: Multiple sources/single buffer scenario (Q1614300)

The paper proposes a neural network based adaptive control methodology to prevent congestion in high-speed asynchronous transfer mode networks. The buffer dynamics at the network switch is considered as a nonlinear process. The adaptive congestion controller scheme involves a one-layer neural network. Tuning laws are provided for the neural network and closed-loop convergence together with stability are proven. It is shown that the performance in terms of cell loss ratio can be reduced to arbitrarily small values by choosing the gains large enough.