Solutions to Parallel and Distributed Computing Problems: Lessons from Biological Sciences

10.6 GSM RADIO RESOURCE MANAGEMENT

In a Global System for Mobile (GSM) system,¹ radio channels must allocate for call setup, handover, and release on a call-by-call basis. GSM provides measurement reports for radio resource management, delivering data about the radio propagation as seen by the mobile station. Currently, these data are used for handover, power control decisions, and measurement reports. These measurement reports constitute a receive-level trace, which is characteristic of the location of the mobile station. Pattern-recognition techniques can be used to classify receive-level traces, thereby improving GSM radio resource management [15].

In [11] and [12], the authors proposed a neural network specifically designed for the level trace pattern-recognition problem. They refer to their system as a Node Splitting Distance Classifying Network (NOSDIC). The network recognises a distance classifier by comparing input pattern vectors to exemplar vectors stored in the network. Input pattern vectors are the received-level traces, smoothed by an input filter. The resulting classification has an error rate of 6% for 68 classes and a varying number of training paterns, i.e., the recognition rate is 94% for the same case and above 90% for all cases simulated so far. However, this scheme is velocity-dependent classification, and requires different classes for different mobility. Junius and Kemmeman [10] have shown that using a basic database of radio measurements, the classification ...