Enabling dynamic spectrum allocation in cognitive radio networks

The last decade has witnessed the proliferation of innovative wireless technologies, such as Wi-Fi, wireless mesh networks, operating in unlicensed bands. Due to the increasing popularity and the wide deployments of these technologies, the unlicensed bands become overcrowded. The wireless devices operating in these bands interfere with each other and hurt the overall performance. To support fast growths of wireless technologies, more spectrums are required. However, as most "prime" spectrum has been allocated, there is no spectrum available to expand these innovative wireless services. Despite the general perception that there is an actual spectral shortage, the recent measurement results released by the FCC (Federal Communications Commission) show that on average only 5% of the spectrum from 30MHz to 30 GHz is used in the US. This indicates that the inefficient spectrum usage is the root cause of the spectral shortage problem. Therefore, this dissertation is focused on improving spectrum utilization and efficiency in tackling the spectral shortage problem to support ever-growing user demands for wireless applications. This dissertation proposes a novel concept of dynamic spectrum allocation, which adaptively divides available spectrum into non-overlapping frequency segments of different bandwidth considering the number of potentially interfering transmissions and the distribution of traffic load in a local environment. The goals are (1) to maximize spectrum efficiency by increasing parallel transmissions and reducing co-channel interferences, and (2) to improve fairness across a network by balancing spectrum assignments. Since existing radio systems offer very limited flexibility, cognitive radios, which can sense and adapt to radio environments, are exploited to support such a dynamic concept. We explore two directions to improve spectrum efficiency by adopting the proposed dynamic allocation concept. First, we build a cognitive wireless system called KNOWS to exploit unoccupied frequencies in the licensed TV bands. KNOWS is a hardware-software platform that includes new radio hardware, a spectrum-aware MAC (medium access control) protocol and an algorithm for implementing the dynamic spectrum allocation. We show that KNOWS accomplishes a remarkable 200% throughput gain over systems based on fixed allocations in common cases. Second, we enhance Wireless LANs (WLANs), the most popular network setting in unlicensed bands, by proposing a dynamic channelization structure and a scalable MAC design. Through analysis and extensive simulations, we show that the new channelization structure and the scalable MAC design improve not only network capacity but per-client fairness by allocating channels of variable width for access points in a WLAN. As a conclusion, we believe that our proposed concept of dynamic spectrum allocation lays down a solid foundation for building systems to efficiently use the invaluable spectrum resource.