||Cognitive radios (CR) have been coined as an emerging and powerful technique to solve the conflict between the limited available spectrum and the inefficiency in the spectrum usage. As one of the most important functionalities, spectrum sensing is highly required to accurately observe the spectrum environment and detect the existence of the primary licensed users so as to avoid harmful interference to them. Once there are some spectrum holes, cognitive users should be properly scheduled to transmit data with the aim of optimizing the performance. This thesis is built upon these two philosophies and aims to shed some lights on the framework design principles for cognitive networks, which will jointly take spectrum sensing and user scheduling into consideration. In this thesis, initial efforts are put on devising an efficient cooperative spectrum sensing approach, which will enhance the performance of spectrum sensing so as to maintain quality of service (QoS) for primary users (PU). To overcome the fading effect of the reporting channel, we propose a cluster-based cooperative spectrum sensing method to improve the sensing performance, where all the secondary users are separated into a few clusters and the most favorable user in each cluster is selected to report to the common receiver. The sensing performance in terms of the false alarm probability and missing detection probability are investigated and compared with traditional protocols. Traditionally, spectrum sensing and user scheduling have long been treated as two sperate problems in the network design, but it is never meant to be a good solution for stimulating cognitive users to perform spectrum sensing, nor does it achieve high spectrum efficiency for energy-constrained systems. In this thesis, we endeavor to design a cross-layer framework for cognitive networks, which will encourage cooperative sensing while guaranteeing fairness among cognitive users and providing high throughput for cognitive systems. We shall present a joint PHY-MAC layer method for narrowband cognitive networks, where the user scheduling is made based on both the instantaneous channel conditions and sensing contribution of each user, so as to exploit the benefits of cooperative spectrum sensing in the PHY layer and the multiuser diversity in the MAC layer. To analyze this framework, a game-theoretical approach is developed, where the payoff function is the cognitive user’s overall power efficiency. Next, an evolutionary game based protocol is proposed to further improve the power efficiency and achieve the Nash equilibrium dynamically based on the behaviors of cognitive users. We shall generalize the cross-layer method and endeavor to design a framework for cognitive OFDMA networks, where two types of scenarios will be examined, namely, homogeneous networks without rate adaptation and heterogeneous networks with rate adaptation. As for the first type of network, a novel matching protocol is explored to guarantee fairness for the sensing users while providing approximately the same spectrum efficiency as the Hungarian method. Afterwards, we shall investigate the network design issues with emphasis on the effects of system parameters over the strategy optimization for each cognitive user. Furthermore, two distributed algorithms will be proposed which can reduce the overhead meanwhile preserving a relatively low throughput loss. As for the second type of network, we shall not only distribute a higher scheduling priority for the sensing users compared to the nonsensing users, but also provide different priorities among different sensing users according to their sensing contributions. Moreover, the scenario with imperfect channel state information (CSI) is taken into account and the performance loss is consequently evaluated.