ML-Aided 2D Indoor Positioning Using Energy Harvesters and Optical Detectors for Self Powered Light-based IoT Sensors

Abstract

Advancements in 6G and IoT sensor networks prioritise sustainability and energy efficiency, with positioning services essential for improved functionality. Light-based IoT (LIoT) systems present a promising solution by achieving energy autonomy through Photovoltaic (PV) energy harvesting and enabling Visible Light Communication (VLC) via indoor luminaries as Optical Access Points (OAP). This research explores the repurposing of energy harvesters at LIoT nodes and detectors at OAPs for positioning and orientation detection in energy-autonomous, battery-free, intermittently operating LIoT sensor networks. We propose potential node and OAP designs, validated by proof-of-concept prototypes, utilising conventional Machine Learning (ML) and Deep Neural Networks (DNN) to enhance localisation. Performance evaluations demonstrate 80% positioning accuracy within 12.5 cm tolerance and 68% orientation prediction accuracy. This approach allows LIoT devices to communicate, harvest energy, and determine their position and orientation using the same illumination from OAPs, underscoring the potential of this LIoT-based system as a sustainable solution for next-generation IoT sensors.