Research On Opportunistic Spectrum Access Technology In Cognitive Radio System

With the rapid growth of wireless service demand, the shortage of current wireless spectrum resources is becoming a severe problem. Cognitive radio (CR) is a technical breakthrough in nowdays wireless communication field. CR is an intelligent system with the ability of realizing dynamic spectrum access (DSA). On the premise of not interfering with primary users, CR can access idle frequency bands which are not currently occupied by primary users, and accordingly relieve the crisis of wireless spectrum shortage. DSA which adopts the overlay approach is known as opportunistic spectrum access (OSA). Aiming at deploying CR in the OSA scenario, the dissertation studies single user OSA with channel correlation, distributed multi-user OSA and secondary user OSA with energy constraint. The major works can be outlined as follows:1. We design the algorithm of single user OSA when there exists correlation among multiple channels. OSA process can be divided into three submodules:channel selection policy, sensor operating point setting policy, and access policy. Under the framework of partially observable markov decision process (POMDP), these policies can be obtained by POMDP solver. The basic factor of resolving POMDP is belief vector. When channel states occupied by primary users evolve independently among different channels, the number of dimensions for belief vector equals to the total number of channels. However, when these channel states evolve dependently (i.e. channel correlation exists), the dimensions of beilief vector will grow exponentially with the total number of channels, whose complexity is undesirable. In the context of channel correlation, We propose access algorithm which reduces the dimensions of belief vector from exponential order to linear order, and constructs a static table restoring transition probabilities of channel correlation. The table is used in the belief vector updating process, making the updated belief vector more precise. Through complexity analysis and simulation, we find that the proposed algorithm can achieve nearly the same throughput performance as the exponential order dimension approach, but with reduced complexity.2. In distributed multi-user OSA environment, we investigate into sharing spectrum opportunities among secondary users. Myopic sensing policy has been verified to be optimal in single user OSA setting. However, when myopic sensing policy is directly applied to multi-user setting, users-synchronization problem (i.e. multiple secondary users select the same channel, bringing about channel contention among different users) will occur, deteriorating the network throughput. Randomized sensing policy can weaken user synchronization, but it will also reduce the throughput of each secondary user. In this dissertation, we propose a domain width changeable randomized sensing policy. Base on randomized sensing policy, we change the number of channels which are available for randomized access (i.e. the width of the random domain), this changing approach can achieve a tradeoff between optimal myopic sensing and users-synchronization avoidance. Simulations show that the proposed sensing policy can achieve higher network throughput compared to the randomized sensing policy over the whole channels.3. In the context of secondary user with energy constraint, OSA design should take into consideration not only the expected access reward, but also saving energy of the secondary user. OSA with energy can also be resolved under the framework of POMDP, but the policy submodules have distinguishing features:the sensing policy considers sleeping mode as a possible kind of channel selecting action, the access policy subdivides the sole access action into access action with different energy consumption levels, while the levels are determined by the channel fading conditions. In the process of solving POMDP, the sensing policy is obtained based on idle probability of the channel, while access policy based on transmission energy level. In this dissertation, we set two fixed thresholds for the idle probability and transmission energy level respectively. By comparing the real time observation values with the fixed thresholds, whether to switch to the sleeping mode and whether to access will be determined. Under the assumption that secondary user's initial energy being equal, simulation results demonstrate that the OSA scheme with two fixed thresholds outperforms the OSA scheme without threshold from the perspective of total expected reward.