Date of Award
2024
Document Type
Open Access Master's Thesis
Degree Name
Master of Science in Electrical and Computer Engineering (MS)
Administrative Home Department
Department of Electrical and Computer Engineering
Advisor 1
Aurenice M. Oliveira
Committee Member 1
K. C. Dukka
Committee Member 2
Anthony Pinar
Abstract
This study addresses the challenge of selecting millimeter Wave (mmWave) beamforming pairs for vehicle-to-infrastructure (V2I) communication, to mitigate latency in highly dynamic vehicular environments. We investigate the use of out-of-band sensor data as side information to model mmWave ray tracing paths and predicting a subset of top-K optimal beamforming pairs for efficient and low-latency searches. Unimodal-Fusion Deep Learning (F-DL) networks was applied to enhance mmWave beamforming process. We started by first investigating the centralized architecture, and then explored a novel distributed architecture through federated learning to minimize resource and latency overheads. The distributed architecture incorporates two biased client selection strategies: MaxLoss and heuristic Multi-Armed Bandit (MAB). This innovative approach streamlines beam selection, improving scalability, robustness and dynamic adaptability.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Subramanian, Abishek, "Application of Fusion based Deep Learning Models to Improve Millimeter Wave Beamforming", Open Access Master's Thesis, Michigan Technological University, 2024.
Included in
Automotive Engineering Commons, Computational Engineering Commons, Systems and Communications Commons