A Machine-Learning Approach for the Exemplar Extraction of mmWave Industrial Wireless Channels

• Published in: IEEE open journal of instrumentation and measurement Vol. 1; pp. 1 - 15
• Main Authors: Kashef, Mohamed, Vouras, Peter, Jones, Robert D., Candell, Richard, Remley, Kate A.
• Format: Journal Article
• Language: English
• Published: IEEE 2022
• Subjects: Apertures; Channel modeling; clustering; Delays; exemplar channel; Feature extraction; Industrial Internet of Things; mmWave channels; NIST; unsupervised learning; Wireless communication; Wireless sensor networks; wireless systems
• ISSN: 2768-7236; 2768-7236
• DOI: 10.1109/OJIM.2022.3181309

Industrial wireless channel modeling is essential for the development of Industrial Internet of Things (IIoT) wireless systems. Moreover, millimeter-wave (mmWave) wireless bands have a high potential to be used for IIoT applications because of their high data-rates and the better applicability of having multiple antennas. As a result, we introduce an exemplar extraction approach to be applied on mmWave wireless channel measurements. A machine learning (ML) clustering scheme is used to divide the measured power-angle-delay-profiles into a number of groups with respect to the angle of arrival. Each of the groups is represented by a power-delay-profile (PDP) exemplar to provide a tractable way for testing and evaluation of mmWave IIoT wireless systems through compactly representing different groups based on their spatial characteristics. Hence, testing of wireless communications equipment can be performed over the exemplars to assess their spatial performance with a significantly reduced amount of data, allowing the development of lab-based device evaluation in a realistic, yet repeatable, test environment. Governing equations are provided in sufficient detail for users to implement the technique in their own labs.