Reserach On Key Technologies Of Dynamic Spectrum Access In Cognitive Radio

With the continuous growth of requirements for wireless communication, spectrum resources have become more and more scarce. However, the fixed spectrum assignment policy hampers the efficient utilization of limited spectrum resources. To solve this contradiction, the concept of cognitive radio(CR) has been introduced, which allows the CR users to access idle licensed spectrums for data communication without incurring harmful interference to the primary users. One of the key technologies of CR is the dynamic spectrum access, which allows the CR users to find appropriate idle licensed channels quickly and effectively for the communications among CR users and, meanwhile, protect the communications of the primary users.By using the theories of optimization, markov decision process and reinforcement learning, the key technologies of dynamic spectrum access in CR systems are investigated in this dissertation: the sensing-transmission optimization by considering the time-varying property of wireless channel; the joint optimization of channel access and transmission rate adaptation by considering the transmission power constraint in CR network; the design of learning scheme for channel access and resource allocation in decentralized CR network. The main contents of this dissertation are summarized as follows:In Chapter 2, we investigate the sensing-transmission optimization by considering the time-varying property of wireless channel. In single-channel environment, we develop the optimal adaptive sensing-transmission schemes for both single-user and multi-user cases, which can adaptively adjust the sensing and transmission strategies according to the time-varying property of wireless channel. For maximizing the throughput of CR users, the adaptive sensing-transmission schemes adjust the sensing-transmission strategies according to the estimation for the current channel state. The adaptive sensing-transmission schemes require no priori knowledge of the ON/OFF activity statistics and the protection for primary users can be guaranteed. We further investigate the sensing-transmission optimization by considering the time-varying property of wireless channel in multi-channel environment. We design the adaptive sensing-transmission schemes in two cases: the channel state information(CSI) is unknown or known by the CR users. The simulation results show that the multi-channel adaptive sensing-transmission schemes can effectively improve the throughput of CR users.In Chapter 3, we investigate the joint optimization of channel access and transmission rate adaptation by considering the transmission power constraint and the time-varying property of wireless channel in CR networks. Various factors, i.e., energy consumption, time-varying nature of wireless channel, spectrum sensing error and transmission collisions, which can influence the performances of dynamic access are considered in the optimal policy design, which aims for maximizing the throughput of CR networks. In centralized environment, the optimal policy can be obtained by solving the constrained markov decision process. In decentralized environment, we prove the existence of the optimal policy and further obtain the optimal policy according to the theory of constrained nash equilibrium. As the complexity of finding the optimal policy increases exponentially with the size of action space and state space, we investigate the complexity reduction from two aspects: firstly, we apply the action set reduction and state aggregation for reducing the complexity of searching the optimal policy; Secondly, we prove that under certain condition, the proposed multi-channel access and transmission rate adaptation policy can be separately solved for each channel. In addition, for the purpose of practicability, the optimal policy is obtained based on the model of pure policy. Finally, we propose the maximum likelihood estimation and first order moment estimation for effectively estimating the ON/OFF parameters in the corresponding environments.In Chapter 4, we investigate the design of learning schemes for channel access and resource allocation in decentralized environments. In this way, even the feedback information is limited, the CR users can still improve the access performances by separately exploiting and learning in the CR environments. Firstly, we investigate the joint design of channel access and transmission rate adaptation for maximizing the transmission energy efficiency in decentralized CR network. From the aspect of cross-layer design, we formulate the utility function by considering both the energy consumption and the QoS performances. In addition, we propose the decentralized access scheme to reduce the information exchange. We also propose the decentralized PD-Q(Probability based Decentralized-Q) learning algorithm and prove the property of convergence. Simulation results show that the PD-Q algorithm can obviously improve the energy efficiency of CR users. Secondly, we investigate the relay transmissions of multiple CR transmit-receive pairs in decentralized environment. Due to the difference of geographical location and available spectrum resource, the transmitter and receiver may not be able to find the same available channel to transmit. To solve this problem, we propose the fully distributed ME-NR(Moment Estimation-No Regret) learning algorithm, which estimates the channel parameters based on the history rewards and adjusts the strategy by using no-regret learning. By using the ME-NR learning algorithm, each transmit-receive pair can independently select the relay node and spectrum resource for maximizing the throughput performance. Simulation results show that the ME-NR algorithm can obviously improve the throughput of CR transmit-receive pairs.