Visible Light and Camera-based Receiver Employing Machine Learning for Indoor Positioning Systems and Data Communications

Younus, Othman Isam (2022) Visible Light and Camera-based Receiver Employing Machine Learning for Indoor Positioning Systems and Data Communications. Doctoral thesis, Northumbria University.

Preview

Text (Doctoral thesis) younus.othman_phd(15037057).pdf - Submitted Version Download (7MB) | Preview

Abstract

Indoor location-based services have played a crucial role in the development of various Internet of Things applications over the last few decades. The use of radio frequency (RF)-based systems in indoor environments suffers from additional interference due to the high penetration rate and reflections of the RF, which may severely affect positioning accuracy. Alternatively, the optical technology using the existing light-emitting diode (LED)-based lights, photodetectors (PDs), and/or image sensors could be utilised to provide indoor positioning with high accuracy. Because of its resilience to electromagnetic interference, license-free operation, large bandwidth, and dual-use for illumination and communication, visible light positioning (VLP) systems have shown great potential in achieving high-precision indoor positioning. This thesis focus is on investigating VLP systems based on employing a single PD, or an array of PDs in the form of a single image sensor (i.e. a camera) for both localization and data communication. Following a comprehensive literature review on VLP, the key challenges in existing positioning methods for achieving a low-cost, accurate, and less complex indoor positioning systems design are highlighted by considering the design characteristics of an indoor environment, position accuracy, number of light-emitting LED, PD, and any additional sensors utilized. The thesis focuses on the major constraints of VLP and provides novel contributions. In most reported VLP schemes, the assumptions of fixed transmitter (Tx) angle and height may not be valid in many physical environments. In this work, the impact of tilting Tx and multipath reflections are investigated. The findings demonstrated that tilting Tx can be beneficial in VLP by leveraging the influence of reflections from both near- and far-walls. It also showed that proposed system offers a significant accuracy improvement by up to ~66% compared with a typical non-tilted Tx VLP system.Furthermore, increasing robustness of image sensor-based receiver (Rx) is a major challenge, which is being addressed using a novel angle of arrival-received signal intensity and a single LED. Experimental results show that the proposed algorithm can achieve a three-dimensional root mean squared error of 7.56 cm. Visible light communications employing a camera-based Rx is best known as optical camera communications (OCC), which can also be used for VLP. However, in OCC the transmission data rate is mainly limited by the exposure time and the frame rate of the camera. In addition, the camera's sampling introduces intersymbol interference Indoor location-based services have played a crucial role in the development of various Internet of Things applications over the last few decades. The use of radio frequency (RF)-based systems in indoor environments suffers from additional interference due to the high penetration rate and reflections of the RF, which may severely affect positioning accuracy. Alternatively, the optical technology using the existing light-emitting diode (LED)-based lights, photodetectors (PDs), and/or image sensors could be utilised to provide indoor positioning with high accuracy. Because of its resilience to electromagnetic interference, license-free operation, large bandwidth, and dual-use for illumination and communication, visible light positioning (VLP) systems have shown great potential in achieving high-precision indoor positioning. This thesis focus is on investigating VLP systems based on employing a single PD, or an array of PDs in the form of a single image sensor (i.e. a camera) for both localization and data communication. Following a comprehensive literature review on VLP, the key challenges in existing positioning methods for achieving a low-cost, accurate, and less complex indoor positioning systems design are highlighted by considering the design characteristics of an indoor environment, position accuracy, number of light-emitting LED, PD, and any additional sensors utilized. The thesis focuses on the major constraints of VLP and provides novel contributions. In most reported VLP schemes, the assumptions of fixed transmitter (Tx) angle and height may not be valid in many physical environments. In this work, the impact of tilting Tx and multipath reflections are investigated. The findings demonstrated that tilting Tx can be beneficial in VLP by leveraging the influence of reflections from both near- and far-walls. It also showed that proposed system offers a significant accuracy improvement by up to ~66% compared with a typical non-tilted Tx VLP system.Furthermore, increasing robustness of image sensor-based receiver (Rx) is a major challenge, which is being addressed using a novel angle of arrival-received signal intensity and a single LED. Experimental results show that the proposed algorithm can achieve a three-dimensional root mean squared error of 7.56 cm. Visible light communications employing a camera-based Rx is best known as optical camera communications (OCC), which can also be used for VLP. However, in OCC the transmission data rate is mainly limited by the exposure time and the frame rate of the camera. In addition, the camera's sampling introduces intersymbol interference.

Actions (login required)

View Item

Downloads