Current radio frequency (RF) spectrum is shown to be highly underutilized in both spatial and temporal domain. One potential solution to improve the spectrum utilization is through cognitive radio (CR). Traditional CR approach adopts the sense-and-transmit strategy to exploit more transmission opportunities. After CR senses the RF channel, it will utilize this channel as long as it is not occupied by licensed users. However, there are some limitations using this approach. First, if the licensed users appear while CR is transmitting on the same channel, the collision occurs. Second, if the licensed users stop transmitting and CR does not detect this opportunity, the throughput loss occurs.
In this dissertation, we focus on studying the statistical properties of the licensed channel bands, and explore the approaches to exploit the traﬃc knowledge to improve the traditional CR strategy. The statistical knowledge of the busy/idle period on the licensed channel allows CR to switch to better channels to increase throughput. To this end, our goal is to develop traﬃc-aware spectrum sharing protocols. As a ﬁrst step, we proposed a traﬃc classiﬁer to classify the distribution for the idle/busy period using less number of observations and extended to the cases where there is no perfect knowledge of traﬃc parameters and periods. We also proposed a traﬃc prediction algorithm based on Markov chain derivation and estimated traﬃc parameters, and we analyzed the impact of estimation errors on the prediction accuracy. Based on this result, we further
designed spectrum sharing protocols with multi-channel sensing strategies for CR
users. The CR users are scheduled for their transmission and cooperative sensing
strategies in order to achieve maximal user throughput. We proposed to solve
the scheduling optimization problem with four low-complexity strategies, i.e.,
sequential, parallel, sequential-parallel, and iterative-parallel sensing strategies.
The optimal sensing order for sequential strategy was derived, and we also pro-
posed several heuristics to further reduce the computation complexity for other
strategies. Furthermore, this protocol is designed to be robust to both channel
and traﬃc uncertainties.