A Comparison of Deep Learning Models for IQ Fingerprint Map Based Indoor Positioning in Ship Environments

Abstract

The importance of indoor positioning has grown in numerous application areas such as emergency response, logistics, and industrial automation. In ships, indoor positioning is also needed to provide services to passengers on board. Due to the complex structure and dynamic nature of ship environments, conventional positioning techniques have limitations in providing accurate positions. Compared to other indoor positioning technologies, Bluetooth 5.1-based indoor positioning technology is highly suitable for ship environments. Bluetooth 5.1 attains centimeter-level positioning accuracy by collecting In-phase and Quadrature (IQ) samples from wireless signals. However, distorted IQ samples can lead to significant errors in the final estimated position. Therefore, we propose an indoor positioning method for ships that utilizes a Deep Neural Network (DNN) combined with IQ fingerprint maps to overcome the challenges associated with accurate location detection within the ship. The results indicate that the accuracy of our proposed method can reach up to 97.76%.