On Capacity Optimization Of Dynamic Spectrum Sharing Systems With Cognitive Radios

This dissertation focuses on the capacity optimization of dynamic spectrum sharing systems with cognitive radios. Beginning from introduction on the essential concepts of cognitive radio technology, it first models and analyzes the dynamics of hierarchi-cal access systems to demonstrate the significance of spectrum sharing for improving radio spectrum efficiency. Then it researches the impacts of multi-user diversity and spectrum sensing on the achievable capacity of spectrum shaing systems, which paves the way for the followed capacity optimization. Further, from optimizing MAC frame structure and designing access algorithm and power control perspectives, it researches capacity optimization of dynamic spectrum sharing systems. The main research con-tents and results are as follows.1) Model the dynamics of hierarchical spectrum sharing systems with Markov Processes and deduce the performance indexes of interest to evaluate and compare the performances of opportunistic spectrum access (Overlay/OSA) and concurrent spec-trum sharing (Underlay). Considering the impacts of both primary users'and secon-dary users'activities on system state, it first models the dynamics of Overlay with an N -demension and 3N -state Markov process and the dynamics of Underlay with an N -demension and 4N -state Markov process. Based on proposed models, it deduces the performance indexes such as block probabilities of primary users and secondary users, channel utilization efficiency and so on as closed forms. With these indexes, the differences between Overlay and Underlay and no-sharing system are demonstrated, which disclose the necessity of spectrum sharing and the significance of modeling system dynamics for performance analysis and optimization.2) Research the impact of multi-user diversity on the capacity of Underlay by mod-eling the system capacity into a function of the number of diversity users and research the impact of spectrum sensing on the achievable capacity of OSA by modeling the throughput of cognitive network into a function of the sensing time. First, considering secondary users diversity (competition) scenario, it models the achievable capacity of Underlay with the number of diversity users, then defines and deduces the multi-user diversity gain, with which the impact of multi-user diversity on achievable capacity of Underlay is researched from theory and simulation perspectives. Second, considering sensing-throughput tradeoff of OSA, it models the throughput of cognitive network with the sensing time, thus the impacts of the sensing time and sensing quality on the achievable throughput of cognitive network are researched. These researches provide some insights for capacity optimnization of spectrum sharing systems.3) Based on analyzing the channel handoff of secondary users with OSA, it jointly optimizes the sensing time and MAC frame duration by balancing the sensing indexes (receiver operating characteristc, ROC) and the sensing-throughput tradeoff of OSA. Aiming at OSA with periodic channel sensing, it first researches generalized channel handoff by modeling the dynamics of OSA to deduce the probabilities of successful handoff, block and no handoff of secondary users as well as the average handoff delay as closed forms. Three application scenarios of MAC frame are found due to handoff, which leads to a new throughput model of cognitive network, a function of both the sensing time and the frame duration. By balancing the sensing indexes and maximiz-ing the throughput of cognitive network, it deduces the optimal sensing time and the optimal frame duration, and the maximum throughput of cognitive network is derived as well. Theoritical and simulated results disclose the impacts of the sensing quality and channel handoff on the achievable throughput of cognitive network.4) By modeling the dynamics of primary network and the spectrum sensing activity of secondary users in OSA system, it defines and deduces a cross-layer performance index as the access reward function to develop a cross-layer optimal access algorithm. Aiming at OSA with periodic channel sensing, by modeling the dynamic of primary network and the spectrum sensing activity of secondry users, it first deduces the prob-ability of secondary user transmiting on each channel without collision, which is the MAC layer index termed no-collision transmission probability. Combined with the physical layer index termed channel capacity, it deduces the achievable capacity of secondary user transmiting without collision on each channel, termed effective trans-mission capacity. Under the criterion of maximizing effective transmission capacity, it develops an optimal cross-layer access algorithm, with which the maximum average effective transmission capacity of OSA is deduced. Theoritical and simulated results indicate that the proposed cross-layer optimal OSA algorithm can provide secondry users with the higher effective transmission capacity than the phyical-layer optimal algorithm and the MAC-layer optimal algorithm.5) Propose the hybrid Overlay/Underlay spectrum sharing scheme and deduce the system capacity model by modeling the dynamics of primry network, then to optimize the capacity with power control. Combining the advantages of Overlay and Underlay, it proposes a hybrid spectrum shaing scheme which permits secondry users automati-cally switch working state from Overlay state to Underlay state with power control according to the dynamics of primary network. Considering the primary network with FDMA and CDMA respectively, it first analyzes the dynamics of primary network to find the time fraction of secondary users at Overlay state and that at Underlay state, which leads to the capacity of hybrid spectrum sharing. Under the criterion of maxi-mizing system capacity, it deduces the optimal transmission power of secondary users at Overlay state and that at Underlay state. The achievable maximum capacity of sec-ondary users is deduced as well. Theoritical and simulated results indicate that hybrid spectrum sharing can provide higher spectrum efficiency than Overlay or Underlay. In summary, above researches optimize the capacities of spectrum sharing systems with cross-layer design and scheduling. The proposed schemes and algorithms benefit to improve dynamic spectrum sharing design. They also provide some insights for fu-ture dynamic radio resource allocation and management.