Ultra Wideband location in scenarios without clear line of sighta practical approach

Abstract

Indoor location has experienced a major boost in recent years. location based services (LBS), which until recently were restricted to outdoor scenarios and the use of GPS, have also been extended into buildings. From large public structures such as airports or hospitals to a multitude of industrial scenarios, LBS has become increasingly present in indoor scenarios. Of the various technologies that can be used to achieve this indoor location, the ones based on ultrawideband (UWB) signals have become ones of the most demanded due primarily to their accuracy in position estimation. Additionally, the appearance in the market of more and more manufacturers and products has lowered the prices of these devices to levels that allow to think about their use for large deployments with a contained budget. By their nature, UWB signals are very resistant to the multi-path phenomenon, so in a situation of good direct vision between two devices (line-of-sight (LOS)) the technology is able to provide very accurate distance estimates. This is not the case when this line of sight is totally or partially blocked (non-line-of-sight (NLOS)). In this case the possible rebounds of the signal can be wrongly confused with the original emitted signal and result in an estimate of time, and therefore distance, very far from the actual value. The aim of this thesis is to offer a complete solution to the problem of indoor location with UWB from a practical approach. This means that the problem is approached from all the angles that should be taken into account in a real use case: basic research in search of an improvement at technical level of the UWB values, prior simulation to ensure a good deployment and performance and the final integration of the solution within an external operational system that allows LBS or clients to make use of location data. The first part of this proposal is presented in this thesis in the first chapters, and consists of using machine learning (ML) techniques to detect and mitigate the possible NLOS measurements generated by low-cost UWB devices. For this task the memory describes the different measurement campaigns with real hardware that were carried out to obtain the training data of the algorithms, and how these were used in different experiments to measure to what extent the proposed solution was valid in a real environment. The second part of the thesis presents data from a location-oriented UWB devices simulator created to perform tests and detect possible problems before deploying real hardware. Among the data presented during this second part, different comparisons between real and simulated scenarios are shown in order to check how reliable the simulation is. In addition, the simulator is presented as part of a case commonly used in indoor location systems: the location of vehicles in an industrial environment using multiple sensors. Finally, the third and final part of this thesis describes a multi-technology real-time location system (RTLS) capable of serving the different actors in a real location system. The platform serves as a link between the location sensors themselves, the LBS and end clients, the researchers in charge of the location algorithms and the managers of derived services such as alerts. By means of the proposed RTLS platform it is possible to decouple the functionality of these entities so that all of them can work without dependencies with the others. In addition to the description of the platform, this memory also includes a summary of the real localization projects in which it has been used, as well as a list of the various modifications and improvements that the platform has undergone in recent times thanks exactly to the experiences obtained after those works on site with real projects.