Estimação do Ângulo de Chegada utilizando Bluetooth 5.1 e redes profundas

Abstract

Indoor positioning systems have gained prominence driven by the increased use of mobile devices and the need for precise location or guidance in various sectors. Bluetooth Low Energy (BLE) 5.1 stands out for spatial orientation tasks due to its low energy consumption and the availability of resources for obtaining the angle of arrival (AoA), through IQ quadrature samples. Processing these samples is critical to AoA investigation. However, the literature lacks databases of these radio signals with real samples, in addition to not yet having studies applying concepts of machine learning and, specifically, deep learning, to estimate the angle of arrival. In this context, this work proposes the use of deep learning techniques to study AoA through a regression model applied to Bluetooth-based indoor guidance systems using real collected samples. We present the entire collection scenario and the procedures necessary to validate the samples and the proposed model. Using the Mean Absolute Error (MAE) performance metric, we found an error of 1.38° in estimating the 135° angle and, overall, the proposed model presented an MAE of 1.87°.