Prediction Based Wireless Resource Management Technologies For Cognitive Radio Networks

To fully satisfy the communication requirements of users, wireless network have developed rapidly. However, many challenges are introduced. On one hand, deployment of wireless networks in different formats meets the growing service needs, but that has led to the increasing serious interconnection problems between heterogeneous networks. On the other hand, the contradiction between spectrum scarcity and inefficient utilization of wireless spectrum becomes increasingly prominent. It is not only because the spectrum resources are limited, but also the traditional spectrum resource allocation strategies are static. Due to its characteristics of flexibility, autonomy and adaptation, Cognitive Radio (CR) has been proposed as an effective solution to the problem of heterogeneous networks convergence, and it can effectively improve the spectrum efficiency of the network through dynamic allocation of spectrum (DAS). Recently, Cognitive Radio Network (CRN), which based on CR, has received a lot of attentions. This thesis focus on the wireless resource management and optimization of CRN, which intents to provide solution to heterogeneity interconnection, improve the spectrum and energy efficiency and balance the traffic load between heterogeneous networks. Specifically, the research contents of this thesis include the following three parts:Firstly, the problem of spectrum sensing in CRN is discussed. An efficient spectrum occupancy prediction scheme based on multilayer feedforward neural network (MFNN) is proposed. As it is difficult for secondary users (SUs) to get priori information from primary users (PUs) in CR system, some models which do not require a priori knowledge of underlying distributions of the observed spectrum occupancy process are needed. Fortunately, neural network (NN) over statistical model offers such an advantage. With the result of prediction, SUs can make a choice on spectrum sensing, which not only helps SUs to save more time for data transmission, but also the detection accuracy will be improved. Consequently, the performance of CR network, such as throughput, can make great improvement. Simulation results in different frequency points are illustrated to validate the prediction accuracy and the generalization ability of the proposed scheme.Secondary, the time-domain resource allocation problem of CRN is considered. In CRN, to efficiently utilize the spectrum resource, one important issue is to allocate the sensing and transmission duration reasonably. In this part, a joint design of energy efficient sensing and transmission durations to maximize EE capacity (EEC) of CRN is proposed. The tradeoff between EEC and sensing and transmission durations are formulized as an optimization problem under constraints on target detection probability of SUs and toleration interference threshold of PUs. To obtain the optimal solution, optimizing sensing duration and transmission duration will be first performed separately. Then, a joint optimization iterative algorithm is proposed to search the optimal pair of sensing and transmission durations. Analytical and simulation results show that there exists a unique duration pair where the EEC is maximized, and that the EEC of the proposed joint optimization algorithm outperforms that of existed algorithms.Lastly, the traffic load balancing problem in CRN is studied. This part aims to solve the problem regarding how to perform offloading efficiently in CRN. Specifically, two schemes have been proposed in support of the offloading process, which are the adaptive offloading scheme and the improved traffic prediction scheme, respectively. The adaptive offloading scheme employs the fuzzy logic algorithm to make offloading decisions. And the traffic prediction based scheme, which adopts the NN model to realize the service loads prediction, serves to enhance the offloading performance. By analyzing the simulation results, it is concluded that our proposed schemes can balance the traffic load between heterogeneous networks and achieve better spectrum efficiency.The thesis summarizes the contents of the full text at the end, and further directions for future research are also discussed.