Impact of LOS/NLOS Transitions on Signal Quality Predictions
Caleb G. Braga, Victor Farias Monteiro, Diego Aguiar Sousa, Tarcisio F. Maciel, Francisco R. P. Cavalcanti

DOI: 10.14209/sbrt.2021.1570731629
Evento: XXXIX Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2021)
Keywords: Channel prediction ARIMA 5G
Abstract
An envisioned characteristic of beyond 5G networks is the regular use of Machine Learning (ML) based solutions in order to enhance the network performance. One possibility is to predict the signal quality in advance in order to anticipate actions, e.g., anticipate signaling related to handover and preallocate resources. One important point that must be investigated is how ML based models are impacted by dynamic changes in the environment. More specifically, the present work investigates the impact of LOS/NLOS transitions on signal quality prediction. The selected algorithm for this study was the well known ARIMA. Simulation results showed that the LOS/NLOS transition is a critical moment for predictions, since previous and future samples are highly uncorrelated. Besides, it was also shown that varying the prediction size window impacts more than varying the user speed.

Download