Thesis Defense - Weingart

Advances in process technology, manufacturing, and architecture have ushered in an age of faster, smaller, and cheaper electronic devices. Emerging processor technology has made it possible to migrate applications that were traditionally implemented in custom silicon to general purpose processors. In the area of wireless communications, this transition has given birth to the field of cognitive and software-defined radio (C/SDR). These C/SDRs offer a broad range of opportunities for improving the use and utilization of radio frequency spectrum. This includes the creation of radio networks that can reconfigure their operation based on application requirements, policy updates, environmental conditions, and the ability to adapt to a wide range of protocols. One of the key benefits of having a C/SDR is its ability to change communication parameters in response to changes in application needs as well as changes in the radio frequency landscape. Such reconfiguration requires an understanding of how these communication parameters interact within the network protocol stack. Analysis of these parametric cross-layer interactions is a critical precursor in the development of a predictive model and algorithm for dynamic reconfiguration of a C/SDR.

This work investigates how parameters at the physical, data link, network, and application layers interact, how desirable configurations of these parameters can be determined, and how they effect the performance of file transfer and Voice over IP (VoIP) applications. An analysis of varying communication parameters across networking layers is used to inform the design, implementation, and evaluation of a predictive model and algorithm for dynamic reconfiguration of a cognitive radio. This model and algorithm allow a C/SDR to dynamically modify its configuration in order to improve system performance. A systematic method for development of a cognitive platform is presented. This method uses statistical analysis of variance (ANOVA) and design of experiments (DOE) techniques to inform the design and implementation of a dynamic reconfiguration algorithm. This algorithm exploits cross-layer interactions to improve system performance, adapt to the needs of users, and respond to changes in the radio frequency environment.