Finding Optimal Action Point for Multi-Stage Spectrum Access in Cognitive Radio Networks

A critical challenge in Cognitive Radio Networks (CRN) is to make decision in real-time on accessing and releasing available channels that maximize the spectrum utilization and the overall system throughput. In this work, we make investigations on optimal action point to explore and exploit the frequency-temporal diversity in addition to spectrum availability. By modeling the Rayleigh fading channel under primary user (PU) activity to be a finite state Markov channel (FSMC) with an absorbing state, we formulate this difficulty into a 2-Dimension optimal stopping problem. Further, we've proved that, the complexity of the 2D optimal stopping rule can be reduced to one threshold policy, where the optimal character still holds. After properly constructing multi-absorbing-states Markov chain for dynamic analysis, we get the throughput of our strategy accurately. Numerical and simulations results have verified that, our threshold based access/switch strategy gains much more throughput than conventional idle/busy based access/switch strategy at the cost of access delay in most cases.