2019/04/22 by Abdelrahman Taha, Taha, Abdelrahman, Muhammad Alrabeiah +3 · 12 citations
Engineering · #Advanced Wireless Communication Technologies #FOS: Computer and information sciences #FOS: Electrical engineering #Indoor and Outdoor Localization Technologies #Information Theory (cs.IT) #Signal Processing (eess.SP) #Underwater Vehicles and Communication Systems #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1904.10136
openalex publication_date 2019/04/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Employing large intelligent surfaces (LISs) is a promising solution for\nimproving the coverage and rate of future wireless systems. These surfaces\ncomprise a massive number of nearly-passive elements that interact with the\nincident signals, for example by reflecting them, in a smart way that improves\nthe wireless system performance. Prior work focused on the design of the LIS\nreflection matrices assuming full knowledge of the channels. Estimating these\nchannels at the LIS, however, is a key challenging problem, and is associated\nwith large training overhead given the massive number of LIS elements. This\npaper proposes efficient solutions for these problems by leveraging tools from\ncompressive sensing and deep learning. First, a novel LIS architecture based on\nsparse channel sensors is proposed. In this architecture, all the LIS elements\nare passive except for a few elements that are active (connected to the\nbaseband of the LIS controller). We then develop two solutions that design the\nLIS reflection matrices with negligible training overhead. In the first\napproach, we leverage compressive sensing tools to construct the channels at\nall the LIS elements from the channels seen only at the active elements. These\nfull channels can then be used to design the LIS reflection matrices with no\ntraining overhead. In the second approach, we develop a deep learning based\nsolution where the LIS learns how to optimally interact with the incident\nsignal given the channels at the active elements, which represent the current\nstate of the environment and transmitter/receiver locations. We show that the\nachievable rates of the proposed compressive sensing and deep learning\nsolutions approach the upper bound, that assumes perfect channel knowledge,\nwith negligible training overhead and with less than 1% of the elements being\nactive.\n