Indoor positioning: investigating the correspondence between continuous position sensors data and manual labeling data

Supervisors: Julien Nembrini

Student: Maxence Feller

Project status: Finished

Year: 2023

Indoor positioning systems are useful tools that can be used in many domains like tracking, space management, business optimisation and many other ones. It is therefore interesting to evaluate the necessary data and methods for their proper functioning. The present document exposes an analysis and experiments conducted in that regard. Its main objective is to explore advantages and disadvantages of position sensors data and manually labeled data. The first part of this work is focused on the placement on a map of a dataset composed of both of these data types. The best solution found for it is to link them by applying a function that matches data from both types together considering recorded position values and timestamps. It also allows to label sensors data according to the manually collected ones. The second part of this work is focused on the improvement of the manual labeling. A web application has been created in order to improve the labeling experience, the quality of the collected data and the accuracy of both positioning and labeling processes. This application has also been evaluated through a user test in order to check its usefulness. The final part of this work discusses the main observations of the whole research. In addition, it proposes a methodology that should improve the quality of all aspects approached in this thesis. The project shows that both data types can be well combined together for positioning and labeling. The conceptualized methodology concentrates the knowledge acquired during this project and opens perspectives for new researches. Continuous position sensors and manual labeling linked together are therefore efficient and still promising for both positioning and labeling processes.

Document: report.pdf