Robustifying Predictive Intelligence for A New Spectrum Market

From autonomous driving to the metaverse, from digital democracy to intelligent healthcare, our next-generation revolutionizing applications are faced with an unprecedented shortage of frequency spectrum resources to meet their high demands from wireless technologies. This calls for a fundamental paradigm shift of the wireless ecosystem, from the current exclusive usage of licensed spectrum to dynamic, market-driven spectrum utilization that allows free spectrum trading and maximizes spectrum efficiency for a better digital future. As a pillar technology of the new spectrum market, intelligent algorithms based on advanced artificial intelligence and machine learning are expected to provide the ever-needed scalability and adaptability for managing and decision making in the new dynamic and complex environment, working in conjunction with or even replacing traditional algorithms used in the current wireless ecosystem. Building on top of the latest advances from machine learning, robust optimization, economic markets and wireless system design, this project designs robust intelligent algorithms for wireless stakeholders to ensure a reliable and resilient wireless infrastructure and ecosystem that can meet the critical requirements of spectrum access for current and future applications. Besides intellectual merit, the research helps develop the future wireless workforce by actively involving and broadening participation from high school and undergraduate students in spectrum-related research.The key innovation of this project is a suite of models and algorithms that fundamentally robustify modeling, optimization, and decision making in dynamic spectrum access using predictive intelligence. Specifically, the proposed research seeks to answer several key questions: how to make sure spectrum management and access is robust against uncertainty from spectrum data and predictive models; how to robustly monitor spectrum activities when spectrum ownership frequently changes; how to ensure a manipulation-free spectrum market with intelligent mechanism design. To answer these questions, the expected contributions of this project are: (1) a suite of techniques and algorithms for quantifying and integrating data and model uncertainty in automated dynamic spectrum access; (2) a new framework to achieve trustworthy spectrum monitoring model re-calibration during spectrum handovers; (3) a set of learning-based spectrum market mechanisms that maximize market efficiency or revenue while being robust against traditional and learning-oriented market manipulation. The outcomes of this research bridge the fundamental gap between the lack of robustness guarantee in current predictive intelligence models and algorithms, and the critical need for robustness in future dynamic spectrum access 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.