Device-free indoor localisation with non-wireless sensing techniques : a thesis by publications presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Electronics and Computer Engineering, Massey University, Albany, New Zealand

Abstract

Global Navigation Satellite Systems provide accurate and reliable outdoor positioning to support a large number of applications across many sectors. Unfortunately, such systems do not operate reliably inside buildings due to the signal degradation caused by the absence of a clear line of sight with the satellites. The past two decades have therefore seen intensive research into the development of Indoor Positioning System (IPS). While considerable progress has been made in the indoor localisation discipline, there is still no widely adopted solution. The proliferation of Internet of Things (IoT) devices within the modern built environment provides an opportunity to localise human subjects by utilising such ubiquitous networked devices. This thesis presents the development, implementation and evaluation of several passive indoor positioning systems using ambient Visible Light Positioning (VLP), capacitive-flooring, and thermopile sensors (low-resolution thermal cameras). These systems position the human subject in a device-free manner (i.e., the subject is not required to be instrumented). The developed systems improve upon the state-of-the-art solutions by offering superior position accuracy whilst also using more robust and generalised test setups. The developed passive VLP system is one of the first reported solutions making use of ambient light to position a moving human subject. The capacitive-floor based system improves upon the accuracy of existing flooring solutions as well as demonstrates the potential for automated fall detection. The system also requires very little calibration, i.e., variations of the environment or subject have very little impact upon it. The thermopile positioning system is also shown to be robust to changes in the environment and subjects. Improvements are made over the current literature by testing across multiple environments and subjects whilst using a robust ground truth system. Finally, advanced machine learning methods were implemented and benchmarked against a thermopile dataset which has been made available for other researchers to use.