IMR: MM-1C: Learning-driven Models for 5G Internet Measurements

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

Enriched with exciting applications providing `smart services,’ next-generation (NextG) cellular networks will fundamentally transform today’s perception of the Internet. Leveraging major advances in artificial intelligence (AI) and machine learning (ML), the envisioned NextG Internet measurement infrastructure will markedly enhance network monitoring, management and deployment. Toward this goal, formidable challenges emerge as measurements can be diverse, complex, and even unavailable. In this context, the present project advocates a learning-driven methodology for systematic, robust, and large-scale NextG Internet measurement infrastructure. This methodology will be informed by extensive, in-depth measurements conducted, and will further utilize the platform and measurement tools developed. Key intellectual advances are pursued in three intertwined research thrusts: T1) novel ensembles of Gaussian Processes to model rich, cross-layer features, and environment, user, and application dynamics; T2) innovative Bayesian active sampling to guide the measurement process; and, T3) path-breaking tools to de-bias, de-noise, and integrate crowd-sourced measurements that may contain missing features (e.g., due to privacy protection); and may be biased, conflicting, or even adversarial. Outcomes will be integrated with an available measurement platform to improve the sampling process, data curation, and analytics that will be continually refined and validated. This research will broadly impact the measurement, design, deployment, and operation of NextG networks, through novel applications and services, many yet to be imagined, thereby bringing significant benefits to the society at large. The project also offers opportunities for engaging undergraduate and K-12 students -- especially women and underrepresented minority groups -- in integrated research, education, and outreach activities. Through collaborations with Industry, the technology transferred will influence the development of NextG networks. In addition, the novel Internet measurement tools will come with open-source software to render the models, codes, datasets, evaluation tests and pertinent artifacts, publicly available. Project’s URL https://spincom.umn.edu/research/currently-funded-projects/nsf2220292 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.