Centralised and Distributed Methods for Dynamic Spectrum Allocation

The evolving market is turning all devices into wireless transceivers, which is leading into a spectrum scarcity and a subsequent search of new and free spectrum pieces. However, detailed analyses have proven that the spectrum is in fact underused and therefore ideas about a more efficient spectrum utilisation are emerging. One of these ideas is Dynamic Spectrum Allocation (DSA) technique. DSA considers the whole spectrum as a common pool of resources to be allocated on demand by means of frequency channel sharing. The whole available spectrum for DSA purposes is split into frequency channels, which will be used by the systems depending on the total number of them required. In the work developed in this Master's Thesis, DSA technique has been applied on top of the primary users accessing via fixed spectrum allocation techniques. This will combine the old spectrum allocation techniques with the new methods, thus defining two types of systems, attending to their spectrum access priority: primary and secondary systems. Two different configurations for the DSA operation have been analysed: a centralised method and a distributed method. In this Thesis, two algorithms have been proposed for the centralised method: in the first one, only the BS is able to measure the frequency channels, and in the second one also the secondary users are able to measure the frequency channels indicated by the BSs. The results indicate that a considerable improvement in the spectrum usage efficiency and the disturbance caused to the primary systems is quite low. In the simulated scenario, the situation when no DSA is applied occupies a 27.07% of the considered spectrum. However, when DSA is used, the spectrum occupied is increased about 60% in the distributed case and about 30% in the centralised case. At the same time, the outage probability caused to primary receivers by the secondary system is lower in the centralised method (under 10%). The distributed method for DSA provides much better results in terms of the spectral efficiency and blocking probability to secondary users (around 5% versus 50% in the centralised methods). However, the centralised method has lower outage probability to primary users (always under 10%). The system proposed to be developed would depend on the performance metrics required and the estimated budget. The centralised methods require less complex algorithms and therefore would be cheaper to develop