Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Real-time localisation system for GPS denied open areas using smart street furniture

Ali, Mohamed (2020) Real-time localisation system for GPS denied open areas using smart street furniture. PhD thesis, Murdoch University.

PDF - Whole Thesis
Download (2MB) | Preview


Real-time measurement of crowd dynamics has been attracting significant interest, as it has many applications including real-time monitoring of emergencies and evacuation plans. To effectively measure crowd behaviour, an accurate estimate for pedestrians’ locations is required. However, estimating pedestrians’ locations is a great challenge especially for open areas with poor Global Positioning System (GPS) signal reception and/or lack of infrastructure to install expensive solutions such as video-based systems.

Street furniture assets such as rubbish bins have become smart, as they have been equipped with low-power sensors. Currently, their role is limited to certain applications such as waste management. We believe that the role of street furniture can be extended to include building real-time localisation systems as street furniture provides excellent coverage across different areas such as parks, streets, homes, universities.

In this thesis, we propose a novel wireless sensor network architecture designed for smart street furniture. We extend the functionality of sensor nodes to act as soft Access Point (AP), sensing Wifi signals received from surrounding Wifi-enabled devices. Our proposed architecture includes a real-time and low-power design for sensor nodes. We attached sensor nodes to rubbish bins located in a busy GPS denied open area at Murdoch University (Perth, Western Australia), known as Bush Court. This enabled us to introduce two unique Wifi-based localisation datasets: the first is the Fingerprint dataset called MurdochBushCourtLoC-FP (MBCLFP) in which four users generated Wifi fingerprints for all available cells in the gridded Bush Court, called Reference Points (RPs), using their smartphones, and the second is the APs dataset called MurdochBushCourtLoC-AP (MBCLAP) that includes auto-generated records received from over 1000 users’ devices.

Finally, we developed a real-time localisation approach based on the two datasets using a four-layer deep learning classifier. The approach includes a light-weight algorithm to label the MBCLAP dataset using the MBCLFP dataset and convert the MBCLAP dataset to be synchronous. With the use of our proposed approach, up to 19% improvement in location prediction is achieved.

Item Type: Thesis (PhD)
Murdoch Affiliation(s): IT, Media and Communications
United Nations SDGs: Goal 9: Industry, Innovation and Infrastructure
Supervisor(s): Koutsakis, Polychronis, Cole, Peter and Oatley, Giles
Item Control Page Item Control Page


Downloads per month over past year