Research On Resource Management In Multihop Cognitive Cellular Networks And Wireless Virtulized Networks With POMDP Formulation

In wireless communications, wireless spectrum is valuable and limited resource. As the user demand for the rate of wireless communications grows significantly, and the applications of wireless communications increases rapidly, the wireless spectrum resource shortage problem is more and more severe. However, the utilization of the existing wireless spectrum resources is very low, of which the utilization efficiency 15%-85%, resulting in the more severe spectrum utilization problem. The disadvantages of traditional long-term fixed spectrum allocation method are obvious. This method is no longer suitable for today's rapidly developing wireless communication technologies. Therefore, some researchers have proposed the Cognitive Radio(CR) technology. The dynamic management of wireless spectrum resources can be realized by CR technology and the utilization efficiency of spectrum can be improved. Besides, the virtualization technology is applied to the wireless networks, which enables an entity to be virtualized into multiple logical resources and allows multiple users to access the resources simultaneously.This thesis mainly studies the dynamic resource management in multihop cognitive cellular networks and virtualized wireless networks. In virtualized wireless networks, the low frequency switching and the joint management problem of cache and spectrum resources is also studied.In multihop cognitive cellular networks, this thesis proposes an information-centric multihop cognitive cellular networks architecture to solve the spectrum resource management problem in information-centric networks. To optimize the system with the objective of maximizing the total reward, the dynamic spectrum management issue in the network is modeled as a Markov decision process with hidden and dynamic parameters, and then this model is modeled as a partially observable Markov decision process to solve the problem. Spectrum resources can be effectively used and the response time of users' requirement can be reduced in this way.In virtualized wireless networks, resource management oriented low frequency switching is considered. To reduce the cost and energy consumption caused by resource switching, the concept of priority is introduced to make the service providers with higher priority use resources continuously. To maximize service providers' profit, the problem of dynamic resource management is modeled as a partially observable Markov decision process. This method can reduce not only the energy consumption caused by resource switching, and also the cost of service providers for higher reward. Furthermore, the joint management of cache resource and spectrum resource is taken into considered. To maximize the total reward of service providers, the cost of cache and the cost of spectrum utilization are considered and the same time and resource management problem is modeled as a Markov decision process with unknown parameter. Then it is modeled as a partially observable Markov decision process. Resource allocation can be more reasonable and the system performance is improved.