CAREER: Over-the-Air Computation - Turning Wireless Networks into Computing Networks

University Of South Carolina At Columbia, Columbia SC

Investigators

Abstract

Wireless signals have long been a cornerstone for many communication-based applications, from mobile cellular systems to navigation. This project aims to use wireless signals in a new direction, i.e., computation, particularly for machines interacting with each other wirelessly, with the ultimate goal of faster computation for fast decision-making, learning, and coordination. To this end, the project plans to use the wireless medium as a computational substrate, i.e., over-the-air computation (OAC), to perform underlying calculations in various applications by harnessing the additive nature of electromagnetic waveforms in the wireless medium. The project plans to answer the fundamental research questions on the reliability, scalability, and feasibility of OAC with three interrelated research thrusts. The foundations of OAC for large-scale systems will be established by designing and advancing (1) reliable OAC schemes that are robust against distortion in wireless channels, impairments, and noise through non-coherent waveforms compatible with continuous functions, new error-correction codes based on modulo-lattice modulation and modulation-on-zeros, and analytical over-the-air computable function expressions based on Kolmogorov-Arnold superposition theorem; (2) scalable OAC frameworks for decentralized optimization, consensus for networked robotics and control systems under latency constraints, and distributed machine learning architectures over wireless networks; (3) methods and procedures that providing insights into the limits of OAC in practice by considering time-frequency-phase synchronization, calibration errors, and the dynamic nature of wireless networks such as mobility. This project’s goal is to improve the understanding of low-latency computation over wireless networks to enable intelligence over wireless systems, real-time learning with the fresh data available at wirelessly connected devices, and faster autonomous systems. The outcomes of this project is expected to be beneficial to multidisciplinary industries such as networked robotics, Internet-of-Things, and vehicular networks. The societal impacts include noticeable latency improvements in computation-heavy applications that affect daily life, ranging from smart cities to environment monitoring and cost reduction through better spectrum utilization. The project integrates research and education through a comprehensive action plan while training engineers and researchers on relevant problems in the areas of communications, computer science, and mathematics. 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.