Modelling Primary System Activity in Dynamic Spectrum Access Networks by Aggregated ON/OFF-Processes

Accurate modelling of primary system activity is key to reliable performance evaluation of dynamic spectrum access (DSA) technologies. Measurements on spectrum use show that in some cases the lengths of the vacant periods in a given frequency band can be correlated, a feature commonly used semi-Markov ON/OFF-models cannot reproduce. In this paper we study the stochastic properties of a simple primary system activity model constructed by aggregating the realisations of several semi-Markov ON/OFF-models. We show that despite its simplicity, our model does result in the desired correlations provided heavy-tailed distributions are used for the period length distributions of the individual processes. We also study in detail the statistical behaviour of our model as parameters such as number of processes being aggregated and distributions of the period lengths are changed. The results show that by suitable parameter selection wide range of behaviours can be modelled, making the model of interest in performance evaluation of algorithms and protocols for DSA networks.