Speaker
Description
Millimeter-wave (mmWave) communications is gaining traction with commercial deployments, but the research community still lacks experimental platforms for prototyping and validating their contributions in real-world settings. Due to the prohibitive cost and complexity of developing Software-defined Radios (SDRs) that natively operate in the mmWave band, a few research efforts explore bridging inexpensive SDR platforms that operate in the sub-6 GHz band with commercial mmWave front-ends that up/down-convert the sub-6 GHz signals to/from the mmWave band, creating programmable mmWave platforms. These platforms mainly focus on using SDRs to control link-level aspects of the mmWave front-ends, e.g., gain, transmission mode, and beam selection, for performing experiments to characterize the mmWave front-ends' irradiation patterns, signal-to-noise ratio, and bit error rate. To the best of our knowledge, no existing open-source programmable mmWave platforms support experimentation with system-level design aspects of mmWave communications such as the initial access (IA), an essential procedure where mmWave radios locate each other and identify the best transmission-reception beam pair to establish communication.
During the IA procedure, mmWave radios iterate over their respective beam codebooks, transmitting reference signals on each probed beam and, by measuring their received power, determining the most suitable beam pair to establish a data link. While the IA procedure is essential to establish and maintain communication between mmWave radios, it also induces substantial delay and decreases the amount of resources available for communication. These overheads scale proportionally to the product of the number of antenna elements on the transmitter and receiver and will become a critical bottleneck as future mobile networks move higher up the frequency spectrum towards THz frequencies, where the number of antenna elements increases to the hundreds. To address this glaring inefficiency of the IA procedure, researchers have proposed many approaches to optimize the procedure by, for example, reducing the number of probed beams. There is a need to develop platforms that allow researchers to experiment and validate their IA methods in over-the-air scenarios.
Recognizing that need, we propose a software-defined mmWave radio framework for experimentation with the IA procedure. We implement our software-defined mmWave radio framework as a GNU Radio Out-of-Tree (OOT) module composed of control and signal processing blocks that serve different functions of the IA procedure, namely: (i) interfacing with the mmWave front-end and controlling the physical hardware; (ii) deciding the beam sweeping set; (iii) collecting and processing metrics that indicate the performance of the current beam pair; and (iv) deciding the best beam pair for data transfer. Such modular design allows experimenters to easily change or replace components of the IA procedure, e.g., the range of beams, sweeping time, and the decision algorithm, and focus only on the relevant implementation aspects of their experimental research.
In this presentation, we introduce our implementation of the software-defined mmWave framework for experimentation on IA. We show how our platform can successfully control a pair of InterDigital's Master Head Unit (MHU) v3 mmWave front-ends through a USRP's GPIO interface to perform reconfigurable and extendable IA, allowing us to evaluate the impact of the range of beams, sweeping time, and the decision algorithms on the IA performance. To facilitate the use of our software-defined mmWave framework by the broader research and development communities, we also define the control logic of the mmWave front-ends via the GPIO interface through platform-dependent configuration files, which can be tailored according to different mmWave front-end models and vendors.
| Talk Length | 15 Minutes |
|---|---|
| Acknowledge | Acknowledge In-Person |
