Wireless indoor localisation within the 5G internet of radio light

Abstract

Numerous applications can be enhanced by accurate and efficient indoor localisation using wireless sensor networks, however trade-offs often exist between these two parameters. In this thesis, realworld and simulation data is used to examine the hybrid millimeter wave and Visible Light Communications (VLC) architecture of the 5G Internet of Radio Light (IoRL) Horizon 2020 project. Consequently, relevant localisation challenges within Visible Light Positioning (VLP) and asynchronous sampling networks are identified, and more accurate and efficient solutions are developed. Currently, VLP relies strongly on the assumed Lambertian properties of light sources. However, in practice, not all lights are Lambertian. To support the widespread deployment of VLC technology in numerous environments, measurements from non-Lambertian sources are analysed to provide new insights into the limitations of existing VLP techniques. Subsequently, a novel VLP calibration technique is proposed, and results indicate a 59% accuracy improvement against existing methods. This solution enables high accuracy centimetre level VLP to be achieved with non- Lambertian sources. Asynchronous sampling of range-based measurements is known to impact localisation performance negatively. Various Asynchronous Sampling Localisation Techniques (ASLT) exist to mitigate these effects. While effective at improving positioning performance, the exact suitability of such solutions is not evident due to their additional processes, subsequent complexity, and increased costs. As such, extensive simulations are conducted to study the effectiveness of ASLT under variable sampling latencies, sensor measurement noise, and target trajectories. Findings highlight the computational demand of existing ASLT and motivate the development of a novel solution. The proposed Kalman Extrapolated Least Squares (KELS) method achieves optimal localisation performance with a significant energy reduction of over 50% when compared to current leading ASLT. The work in this thesis demonstrates both the capability for high performance VLP from non- Lambertian sources as well as the potential for energy efficient localisation for sequentially sampled range measurements.