Dynamic re-configuration in 3G networks based on adaptive time/frequency filtering

In this paper, we present a set of mathematical algorithms capable of determining an efficient load-sharing strategy operating on the basis of the throughput requests received by the single nodes of a 3G network. The main idea of our proposed work is to, fully, exploit the variable geometry of the links and their adaptive filtering capabilities to determine the likelihood of bottleneck congestion and prevent it from happening, before it occurs. Adaptive filtering and links re-configurability play a key role in order to avoid congestion and obtain higher throughput on each link in the network. In order to determine the strategy of preference we use random multi-resolution harmonic analysis techniques. These allow for the rapid analysis of the characteristics of the entire traffic flow through the network with a minimal computational load. The information provided by the algorithms is passed on to different nodes for implementation via filtering and receiver reconfigurations. Each data stream is analyzed as a signal processed through a highly non-linear filter: the time-course of a network throughput displays different features depending on the network-filter impulse response. The detection of these features provides the network control with a solid probabilistic criterion to direct its reconfiguration strategy.