Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
With the prevalence of smartphones, watches, and Internet of Things (IoT) devices, the abilityto track their positions is becoming increasingly important. For many indoor positioningsystems (IPSs), providing an uninterrupted flow of information in real-time, effective, and lowpower are the main criterion. As a result, a solution is needed to withstand these conditions.This study is conducted to fill this gap and propose new solutions for an old question.In this thesis, three main indoor positioning systems are proposed. Primarily, the focus is oncreating an IPS based on a smart-card functioning at 125 kHz. This system is a proximityactivatedcard with a magnetic transmitter that is designed to transmit data to a server. Thestudy revealed an accuracy of 30 cm in 70% of the observations by a sigmoid function-activatedneural network.Another positioning system is studied by introducing a non-blocking access control scenario.This system is a hybrid based on a magnetic positioning system and a visual systemfor thesurveillance of the area. Using a Global Nearest Neighbor algorithm, a merging algorithmis developed and implemented to allocate the data collected from themagnetic system andthe camera to a certain person. This system reached an accuracy of 97% in various testingscenarios.Despite the accuracy of these systems, magnetic-based systems are impossible to integratewith smartphones and other dominant radio frequency systems such as WLAN. While havingno multipath effect and functioning in certain difficult environments, such as next to brickwalls or corners, these systems are prone to huge metallic bodies. Therefore, a new system isintroduced based on Bluetooth Direction Finding (BLE-DF). This technology utilizes seamlessintegration with numerous IoT and smart devices, offering swift performance and low powerconsumption while also boasting highly precise direction-finding capabilities.This thesis studies two main BLE-DF architectures. The design characteristics of the designedantenna arrays, a signalmodel for the arrays, and the BLE-DF are proposed in chapters 4 to 5.The performance of the arrays in an uncontrolled environment revealed that our proposedantenna arrays are up to 88% more accurate than an out-of-the-shelf array. Different angleestimation algorithms and a data modification process are proposed which caused up to 40%increase in precision. Different structures of antennas and patterns are studied. A precisionof 3.7⊠is achieved using a URA array that is equivalent to 30 cm in a range of 5m. Finally, apositioning algorithm is presented that proves the same accuracy of approximately 30 cm.Such a high accuracy enables new applications.While the achieved accuracy is very promising, like any other technology, the BLE-DF has its own limitations. For instance, the system is prone to multipath which is inevitable in indoorspaces. Therefore, several methods are proposed to compensate for this effect, such as usingGaussianMixtureModels (GMM) and spatial smoothing that could increase precision by 85%and, 40% respectively. Finally, this dissertation tries to smooth the path of future interestedresearchers by contributing to a better understanding of the performance of our proposedsystems as low-power, sub-meter accuracy IPSs.
Jan Skaloud, Gabriel François Laupré
,
Yves Perriard, Alexis Boegli, Pooneh Mohaghegh, Rabia Saeed