Backscattered Visual Light Positioning using LED Sensing in Combination with Edge Computing based Artificial Neural Network Classification

Abstract

Indoor Positioning Systems gained notable interest in the recent years. Amongst the various applied technologies and methods, positioning systems that utilize the visible light spectrum, have proven to be ideal candidates to establish such systems. Amongst other embodiments, visible light based systems can be realized by applying backscattered scenarios, where only reflections from the object of interest are utilized to infer its position. This backscattering approach has the advantage that the light source and the photosensitive device can be integrated into the same module. In this work we combine such a backscattering approach with the innovative approach in order to utilize LEDs not only as light sources, but also as receiving elements. Our novel approach demonstrates a sophisticated control mechanism, that uses 4 LEDs as light sources and receiving elements, without any disturbance in the provided illumination being perceivable to an observer. By utilizing an Artificial Neural Network, we perform the task of position determination on the edge, demonstrating the great potential of our solution approach. We verify our approach by experimental results, showing that the position (quadrant) in which a retroreflective foil is located, can be determined correctly with more than 92% accuracy, thus showing the advanced implementation of a highly integrated smart system.