DeepBeam: Deep waveform learning for coordination-free beam management in mmWave networks

  • * Register (or log in) to the Neural Network to add this session to your agenda or watch the replay

  • Date
    5 October 2021
    Timeframe
    17:00 - 18:00
    Duration
    60 minutes
    Share this session

    Because of their highly directional transmissions, radios operating at millimeter wave (mmWave) frequencies need to perform beam management to establish and maintain reliable mmWave links. To achieve this objective, existing solutions mostly rely on explicit coordination between the transmitter (TX) and the receiver (RX), which significantly reduces the airtime available for communication and further complicates the network protocol design. Advances of the state of the art are presented in DeepBeam, a framework for beam management in mmWave networks that does not require pilot sequences from the transmitter, nor any beam sweeping or synchronization from the receiver. This is achieved by inferring (i) the Angle of Arrival (AoA) of the beam and (ii) the actual beam being used by the transmitter through waveform-level deep learning on ongoing transmissions between the TX to other receivers. In this way, the receiver can associate Signal-to-Noise-Ratio (SNR) levels to beams without explicit coordination with the transmitter. To demonstrate the generality of the proposed approach, we conduct an extensive experimental data collection campaign where we collect more than 4 TB of mmWave waveforms with (i) 4 phased array antennas at 60.48 GHz, (ii) 2 codebooks containing 24 one-dimensional beams and 12 two-dimensional beams; (iii) 3 receiver gains; (iv) 3 different AoAs; (v) multiple TX and RX locations. Moreover, we collect waveform data with two custom-designed mmWave software-defined radios with fully-digital beamforming architectures at 58 GHz. We also implement our learning models in FPGA to evaluate latency performance. Results show that DeepBeam (i) achieves accuracy of up to 96%, 84% and 77% with a 5-beam, 12-beam and 24-beam codebook, respectively; (ii) reduces latency by up to 7x with respect to the 5G NR initial beam sweep in a default configuration and with a 12-beam codebook. The webinar will also describe how to access to the waveform dataset and the full DeepBeam code repo. This talk is based on a paper (co-authored with Francesco Restuccia and Tommaso Melodia) presented at ACM MobiHoc 2021.

    Share this session
    Discover more from programme stream
    AI/ML Challenge Finale: Beam-Level Traffic Forecasting and Site specific radio...

    AI/ML Challenge Finale: Beam-Level Traffic Forecasting and Site specific radio...

    13 December 2024 - 14:00 to 15:30

    AI/ML Challenge Finale: Radio Resource Management for 6G and Multi-user...

    AI/ML Challenge Finale: Radio Resource Management for 6G and Multi-user...

    8 November 2024 - 15:00 to 16:15

    Green Energy and Traffic Forecasting using AI/ML Challenge Launch

    Green Energy and Traffic Forecasting using AI/ML Challenge Launch

    6 August 2024 - 14:00 to 15:00

    Enabling Distributed Applications with Online Machine Learning

    Enabling Distributed Applications with Online Machine Learning

    17 July 2024 - 17:00 to 18:00

    Radio Resource Management for 6G in-X Subnetworks

    Radio Resource Management for 6G in-X Subnetworks

    26 June 2024 - 14:00 to 15:00

    Are you sure you want to remove this speaker?