Waveform Design for Over-the-Air Computing

The study on over-the-air (OTA) computing, published on arXiv, addresses the growing challenges posed by the anticipated surge in devices within next-generation networks. By leveraging simultaneous uncoded transmission across multiple access channels, OTA computing promises efficient resource management. The research focuses on critical issues such as transmitter synchronization errors and intersymbol interference, which can hinder performance. Through a theoretical analysis, the authors explore methods to minimize mean squared error (MSE) in OTA transmission and derive optimal power policies using alternating optimization. A significant innovation is the introduction of a deep neural network (DNN)-based approach for waveform design, which aims to outperform traditional waveforms like raised cosine. Simulation results validate the theoretical findings, demonstrating performance gains that are essential for the future of digital communication.