2020 · 65 citations · 10 references
EngineeringComputer ScienceElectrical EngineeringAntennasMachine LearningWireless CommunicationsBeam SteeringBeamformingDeep LearningMillimeter-wave CommunicationsPattern RecognitionRepresentation LearningBeam AlignmentBeam SelectionBeam TrainingBeamforming AlgorithmsBeamforming GainDeep Learning For LocalizationDeep Learning-based LocalizationOutage Probability
Location-aided beam alignment methods exploit the user location and prior knowledge of the propagation environment to identify the beam directions that are more likely to maximize the beamforming gain, allowing for a reduction of the beam training overhead. They have been especially popular for vehicle to everything (V2X) applications where the receive array orientation is approximately constant for each considered location, but are not directly applicable to pedestrian applications with arbitrary orientation of the user handset. This paper proposes a deep neural network based beam selection method that leverages position and orientation of the receiver to recommend a shortlist of the best beam pairs, thus significantly reducing the alignment overhead. Moreover, we use multi-labeled classification to not only capture the beam pair with highest received strength but also enrich the neural network with information of alternative beam pairs with high received signal strength, providing robustness against blockage. Simulation results show the better performance of the proposed method compared to a generalization of the inverse fingerprinting algorithm in terms of the misalignment and outage probabilities.
10
Dropout: a simple way to prevent neural networks from overfitting
Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky et al. · 2014 · 34.3K citations
MmWave Vehicular Beam Selection With Situational Awareness Using Machine Learning
Yuyang Wang, Aldebaro Klautau, Mónica Ribero et al. · IEEE Access · 2019 · 118 citations · Full text