Collaborative Research:SWIFT: Exploiting Application Semantics in Intelligent Cross-Layer Design to Enhance End-to-End Spectrum Efficiency

University Of California-San Diego, La Jolla CA

Investigators

Abstract

The growing demands for network capacity have spurred existing wireless networks to expand into millimeter-wave radio bands with enormous spectrum resources. Theoretically, wireless link capacity grows linearly with the amount of spectrum. Unfortunately, the capacity gain does not straightforwardly translate into improvement of application-layer quality of experience (QoE). This project advocates for the key notion of end-to-end spectrum efficiency and argues that it is imperative to intelligently exploit application semantics to achieve high end-to-end spectrum efficiency. The basic idea is that wireless networks should be made aware of the “utility” of application data when allocating radio resources, while applications should refactor data to better expose and encode application utility and service requirements for end-to-end spectrum efficiency. This project will contribute to a new paradigm for re-architecturing 5G and future wireless networks to support and enable a wide range of innovative future applications, many yet to be imagined, thereby bringing significant benefits to society at large. The outcomes from this research project will be incorporated into the academic curriculum to equip the workforce with the skills needed to develop future wireless networks. The project team will actively recruit, engage, and mentor a diverse group of undergraduate students and budding researchers, with an emphasis on broadening participation by under-represented groups in advanced wireless and computing research. This project advances a vertically integrated, machine-learning-guided, intelligent cross-layer framework to exploit application semantics for end-to-end spectrum efficiency. It aims to re-architect the radio network protocol stack for 5G & beyond networks through fine-grained refactoring of application data in accordance with application semantics and service needs. The proposed framework is designed based on three key principles: 1) exploiting application semantics and data refactoring, so that the radio networks can intelligently allocate the heterogeneous radio resources with different reliability-efficiency properties to match the utility of data; 2) adopting learning as a guiding principle in wireless system design and integrating learning-based methods across the entire network stack, instead of piecewise application; 3) incorporating intelligent real-time decision mechanisms to mitigate the long-tail performance of machine learning methods. This project will spur the broader research community and industry in exploring new directions in enhancing spectrum efficiency from an end-to-end application/service-centric perspective. It will produce open-source hardware, software, and datasets. The project team will engage students at all levels for integrated research and education activities, and will contribute to the Broadening Participation in Computing programs in the PIs' institutions. 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.