We study cognitive wireless networks from a cross-layer perspective, where we investigate the effects of the PHY layer parameters and enhancements on the MAC layer performance. We quantify the benefit of using sophisticated techniques such as cooperative communications and network coding in cognitive networks.

The first part deals with unicast scenarios. We first study the problem of random access over time varying channels with cognitive nodes adjusting their access probabilities according to the decentralized channel state information they acquire at the PHY layer. We derive the conditions for our random access scheme to outperform orthogonal access.

We then study the case where a set of secondary users (SUs) opportunistically accesses the primary user's (PU) spectrum whenever it is idle. Since sensing errors are unavoidable, we study the effect of the interference from the SUs on the stable throughput of the PU. We then compute the range of the SUs' transmission parameters that guarantees the stability of the PU queue. In order to balance the negative effects of the interference from the SUs, we propose a PHY layer relaying protocol between the PU and SU networks that is based on distributed orthogonal space-time block codes. Under this protocol, it is shown that the PU's throughput gain from relaying increases with the number of SUs. Moreover, the SUs might benefit from relaying the PU's packets as well.

Next, we propose and analyze access schemes at the SUs aiming at exploiting the SU's knowledge of the statistics of various channels and of the average arrival rate to the PU. The motivation is that although the traditional opportunistic spectrum access (OSA) guarantees full protection to the PUs, it is sometimes too conservative if the interference caused by the SUs at the PU receiver is negligible. We derive the conditions under which schemes without sensing outperform schemes with sensing since they offer to the SU more data transmission duration.

The second part of the dissertation deals with cognitive multicasting networks. First, we study relay assisted multicasting. The relay delivers the unsuccessful packets of the source during the idle slots of the source which are determined by sensing. This avoids allocating any explicit resources to the relay. We then substantiate the benefit of using network coding (NC) at the relay.

Finally, we study the problem of reliable spectrum sensing and opportunistic access on channels with stochastic traffic in batch processing systems such as NC. We show how an SU can leverage the structure induced by block-based NC on PUs' channels to mitigate the effects of channel sensing errors and improve the throughput. We consider two different objectives at the SU: quickest detection of an idle slot and throughput maximization. We validate our results with real radio measurements taken in software-defined radio based wireless network tests.