Collaborative Research: SWIFT-SAT: INtegrated Testbed Ensuring Resilient Active/Passive CoexisTence (INTERACT): End-to-End Learning-Based Interference Mitigation for Radiometers

Mississippi State University, Mississippi State MS

Investigators

Abstract

As next-generation communication and satellite systems utilize more frequency bands, the potential interference risks to passive radiometer sensors used for environmental and atmospheric sensing are increasing. Thus, it is imperative to develop efficient methods to detect, characterize and mitigate anthropogenic sources of interference at passive radiometers. Radio frequency (RF) research domains, specifically those addressing the active/passive coexistence, are in critical need of datasets that enable learning-based detection, identification, and classification, as was observed in image processing domains. The goals of the project INTERACT (INtegrated Testbed Ensuring Resilient Active/Passive CoexisTence) are 1) to collect/to currate active/passive RF coexistence datasets with ground truth information 2) to develop data-driven learning-based RF interference (RFI) detection and mitigation approaches enabled by the generated data. The datasets will be collected by an airborne passive microwave radiometer system to be deployed on the NSF's AERPAW (Aerial Experimentation and Research Platform for Advanced Wireless) platform. The proposed research will further our undertanding on spectrum sharing through passive sensing methods, RF datasets, and learning based RFI mitigation approaches. The project INTERACT proposes three key innovations: 1) A new Unmanned Aerial System (UAS) based passive radiometer system will be developed. This system together with the experimental development of various active transmission scenarios covering different geometries, transmitter parameters and waveforms at non-restricted bands will result in the first-ever large experimental RF dataset with ground truth information for passive/active RF coexistence. A digital twin for passive radiometry in the emulation environment of AERPAW will be developed to enable experimenters to facilitate extensive, yet realistic RF mitigation experiments in a Cloud environment. 2) Novel data-driven end-to-end learning-based RFI detection and mitigation approaches will be developed. The proposed solutions will focus on approaches that can achieve high-resolution RFI detection in the time-frequency domains, learning based radiometer calibration, and joint mitigation to estimate the scientific observation of radiometers under RFI. These solutions do not require centralized servers and are designed to work on passive radiometer systems in order to detect and mitigate RFI without any information exchange between coexisting systems. 3) The research will produce new deep reinforcement learning and subspace-based RFI mitigation approaches using the feedback from active and passive systems. 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.