Due to the existence of fast fading and co-channel interference in wireless communications systems, the problems of adaptive equalization and interference suppression call for solutions that are beyond the reach of conventional schemes, e.g., those based on least squares and least-mean-squares. This research project is applying novel adaptive signal processing techniques, referred to as Set- Membership Adaptive Recursive Techniques (SMART), to resolve critical problems that arise in advanced communications systems. The novelty of SMART stems from the following: (1) Their outcomes always satisfy a prescribed instantaneous bound on the magnitude of the estimation error, which in adaptive equalization, is the difference between the desired output and the filtered output. (2) They are characterized through a feasible set of parameters, which meet the specification without requiring the existence of a true parameter. (3) They update parameter estimates selectively depending on the innovation of the input data. This feature leads to an adaptor sharing paradigm that significantly improves the cost-effectiveness and data usage efficiency in adaptive signal processing. In simulation studies, SMART have exhibited superior performance and lower computational complexity in applications such as adaptive equalization and interference cancellation. More generally, this project will establish a sound theoretical foundation for Set-Membership Adaptive Filtering (SMAF), and explore the use of SMART in those applications where performance requirements may exceed what traditional algorithms can offer, e.g., modern multiple access communication systems. We anticipate that this research will advance the state-of-the-art in adaptive filtering, both theoretically and practically, and significantly enhance the performance of modern wireless communication systems.
