Deep-Learning-Based Antenna Alignment Prediction for Mobile Indoor Communication
Árpád László Makara,
Botond Tamás Csathó,
András Rácz,
Tamás Borsos,
László Csurgai-Horváth,
Bálint Péter Horváth
Affiliations
Árpád László Makara
Department of Broadband Infocommunications and Electromagnetic Theory, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Műegyetem rkp. 3., H-1111 Budapest, Hungary
Botond Tamás Csathó
Department of Broadband Infocommunications and Electromagnetic Theory, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Műegyetem rkp. 3., H-1111 Budapest, Hungary
András Rácz
Ericsson Research, H-1117 Budapest, Hungary
Tamás Borsos
Ericsson Research, H-1117 Budapest, Hungary
László Csurgai-Horváth
Department of Broadband Infocommunications and Electromagnetic Theory, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Műegyetem rkp. 3., H-1111 Budapest, Hungary
Bálint Péter Horváth
Department of Broadband Infocommunications and Electromagnetic Theory, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Műegyetem rkp. 3., H-1111 Budapest, Hungary
A significant innovation for future indoor wireless networks is the use of the mmWave frequency band. However, an important challenge comes from the restricted propagation conditions in this band, which necessitates the use of beamforming and associated beam management procedures, including, for instance, beam tracking or beam prediction. A possible solution to the beam management problem is to use artificial-intelligence-based procedures to learn the hidden spatial propagation patterns of the channel and to use this knowledge to predict the best beam directions. In this paper, we present a deep-neural-network-based method that has memory that can be used to predict the best reception directions for moving users. The best direction is the highest expected signal level at the next moment. The resulting method allows for a user-side antenna management system. The result was evaluated using three different metrics, thus detailing not only its predictive ability, but also its usability.
