Network selection strategy for cognitive radios with heterogeneous primary networks

Dynamic spectrum sharing is a promising approach to reusing the underutilized radio spectrum in a cognitive radio system (CRS). Potential available spectrum for CRS can be from multiple primary radio systems (PRSs) with different spectrum prices, and secondary users (SUs) with heterogenous usage modes, interference punishments and service types. Choosing which primary network to access for SU brings the network selection problem. This scenario generates new opportunities for CRS to improve the system performance by exploiting heterogenous characteristics of PRSs. In this paper, we first use continuous time Markov chain (CTMC) model to describe the network selection problem and derive the performance metrics. Then we intuitively propose Random and Greedy network selection strategies under this framework. Motivated by the drawbacks of those strategies, we categorize the service types of SU into Real-Time Services (RTS) and Non-Real-Time Services (NRTS) and propose an optimal network selection algorithm to maximize performance metrics based on policy iterations. Lastly, the performance of the proposed designs is compared and evaluated through simulation and numerical results.