Enhanced 2-Step DNN Model for RSSI-Based Indoor Localization

Abstract

In this paper, we introduce a novel approach, termed the 2-Step Deep Neural Network (DNN), designed specifically for indoor localization utilizing Received Signal Strength Indicator (RSSI) data. This method represents an advancement over the previously proposed 2-Step Extreme Gradient Boosting (XGBoost), aiming to enhance estimation precision by leveraging a single coordinate (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$x$</tex> or <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$y$</tex>) as a feature. The pivotal alterations encompass transitioning from the XGBoost to DNN and refining the training data to develop a resilient learning model for positional coordinates. Through comprehensive simulations, we demonstrate that the proposed 2-Step Robust DNN attains superior estimation accuracy while preserving the absence of constraints on the dataset.