Efficient Power-Aware Multi-Channel Selection for Dynamic Cognitive Radio Networks

In dynamic cognitive radio networks (CRNs), unlicensed secondary users (SUs) opportunistically transmit in the spectrum holes left available by licensed primary users (PUs), without knowing in advance their presence on the channels. For SUs with multi-channel transmission capability, the process of selecting the channels to use by sequentially sensing the presence of PUs and evaluating the channel gains, incurs performance tradeoffs, power constraints, and expensive computations. In this paper, the channel selection problem for SU throughput maximization, limited by the power available at SUs, is formulated in a CRN with both heterogeneous PU-traffic and channel conditions. Our proposed optimal algorithm only requires linear time and quadratic-space online complexities and cubic offline complexities, with the latter needed when CRN characteristics significantly change. This represents a significant improvement over existing algorithms. Our solution achieves throughput and delays close to maximum, not compromising accuracy for efficiency as approximate channel selection algorithms need to do.