SWIFT: SMALL: Interference Canceling for Improved Coexistence Between Passive and Active Radio Systems

A difficult problem in the management of the limited radio spectrum is that active (transmitting) users of the radio spectrum must coexist with passive (receive-only) users of the radio spectrum; in particular, with radio astronomy and geophysical remote sensing. This project investigates the prospects for improving coexistence using "canceling". Canceling consists of identifying and subtracting an undesired signal such that the receiver is able to access signals of interest that would otherwise be obscured.

In principle, this allows enhanced sharing of spectrum. However, existing canceler technology is generally not sufficiently reliable for general use. This project aims to improve the performance of cancelers using enhanced signal processing techniques including machine learning.

Success in this effort will (1) improve the ability of active and passive users to share spectrum and to operate in adjacent frequency bands while (2) improving the quality of signals collected in passive applications who must operate in or near bands allocated to active services. The project will engage undergraduate engineering students as researchers through two-semester senior design projects.

Public outreach will include a public-facing website and a YouTube channel documenting the various activities of the project.

The research strategy employed in this project is to address the two principal difficulties that have been found to limit the performance of canceling for passive users in the past, namely (1) that interference can be simultaneously damaging yet too weak to accurately estimate, and (2) interfering signals vary significantly over the timescales over which they must be canceled. The "too weak to estimate" problem will be addressed using concurrent partial cross-layer knowledge of the signal, machine learning, and terrestrial source-sited receivers.

The "dynamic channel" problem will be addressed using augmented physical models and machine learning to improve tracking of parameters. Enhanced cancelers are expected to be able to better mitigate the spectral sidelobes of interferers whose primary emission is located out-of-band relative to the frequency of observation. Further, machine learning will be used to dynamically predict performance and recommend canceling strategies tailored to user- and application-specific objectives.

Performance will be evaluated through the use of simulations, bona fide data collected from operational instruments, and demonstrations using a purpose-built small-aperture telescope testbed located at Virginia Tech.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.