Description
A major shift in wireless communications is now emerging with the development of cognitive radios, which attempt to share spectrum in a fundamentally new way. Cognitive radios address the problem of poor spectrum utilization exhibited in many frequency bands. On a conceptual level, cognitive radio networks sense the spectral environment and adapt transmission parameters to dynamically reuse available spectrum. The novelty of this approach requires us to re-architect the mechanisms for using radio frequencies and find a way for multiple systems to co-exist through sharing rather than fixed allocations.
This dissertation explores some fundamental questions in cognitive radio system design by bridging the theoretical and practical aspects of the physical and network layers. A system level design involved a closed loop research approach connecting theoretical analysis and development of new algorithms with their implementation and experimental verification. A wireless testbed platform with capabilities for real-time signal processing and protocols, networking and multiple antenna communication was developed to support this research approach.
Spectrum sensing has been identified as a key enabling functionality for cognitive radios, therefore the goal of this research was to address its feasibility, performance limits and implementation issues. A major challenge in the spectrum sensing design is the requirement to detect very weak signals of different types in a minimum time with high reliability. To solve this problem, a cross-layer design approach was applied involving sensing radio front-end, digital signal processing and networking solutions. By exploiting spatial filtering for interference suppression, statistical signal processing to combat channels uncertainties and network cooperations to improve detection reliability, we show the practically achievable limits for sensing weak signals in wideband cognitive radio channels. As a result, an architecture of a spectrum sensing function is proposed and its performance and implementation complexity are characterized using the developed testbed platform.