ITR: Development of a Novel Short-Data-Record Adaptive Filtering Framework for Rapidly Changing Communications Environments
Stella Batalama Principal Investigator
MetadataShow full item record
PIs: S. N. Batalama and D. A. Pados<br/><br/>PROJECT ABSTRACT<br/><br/>The effectiveness of a receiver designed for a rapidly changing multiple access communications environment depends on the following design attributes: (i) low computational complexity, (ii) multiple-access-interference resistance, and (iii) system adaptivity with superior performance under limited (short) data support. Adaptive short-data-record designs appear as the natural next step <br/>for a matured discipline that has extensively addressed the first two, (i) and (ii), design objectives in ideal setups (perfectly known or asymptotically estimated statistical properties). System adaptivity based on short data records is necessary for the development of practical adaptive receivers that exhibit superior signal-to-interference-plus-noise ratio (SINR) or bit-error-rate (BER) performance when they operate in rapidly changing communications environments that limit substantially <br/>the input data support that is available for adaptation and redesign.<br/><br/>A novel line of research is identified and pursued in this project that lies in a multidisciplinary intersection of Estimation Theory, Communications Theory, and Mean-Square optimum linear filtering. Consider an arbitrary input signal vector space and a given information bearing signal vector to be recovered in the presence of multiuser or other forms of heavy interference. A key for the successful solution to the problem of adaptive receiver design under short data records is to employ receiver estimators with varying bias/variance characteristics and to control effectively <br/>these characteristics in a data-centric manner. In this project, the investigators develop a short-data-record adaptive filtering framework that involves (i) the tools for the generation<br/>of a sequence of filter estimators with varying bias/variance tradeoff, as well as (ii) the tools for selecting the most appropriate estimator in the sequence for a given input data<br/>record. The Auxiliary-Vector (AV) filtering means developed by the investigators during the past few years as a state-of-the-art practical engineering solution to the problem of short-data-record <br/>adaptive filtering is used as a benchmark case-study. The theoretical and practical implications of such a research framework are far reaching. Biased estimators and algorithms that<br/>offer full control over the bias/variance balance are rarely reported in the literature, if any in a communications applicable context. While the target applications of this work are all critical<br/>communications problems, the pertinent theoretical developments may touch many aspects of multidisciplinary engineering that are hampered by the ``curse of dimensionality'' and could benefit from adaptive filtering and/or adaptive system optimization through limited input data.