Real time tracking and visualization of indoor social interactions

Teachers: 
Denis Lalanne
Student: 
Thomas Rouvinez
Project status: 
Finished
Year: 
2015

During this Master Thesis, we analysed the current state of research in the field of indoor tracking systems technologies. We studied their respective applications and issued a set of criteria that corresponds to the requirements of tracking social interactions. An indoor positioning system based on passive RFID tags carried by the participants of social events turned out to be the best approach. A complete indoor positioning system is then designed, including using RFID readers, capturing participants locations, manage and store the data, retrieve clean data, displaying this data in real time. A multi-tier system is created, combining RFID management and data capture, a web service to interface between the raw data and the final clients exploiting this data. The web service implements a set of methods to process the data, for both online and offline needs.

This project regroups the developments of multiple pieces of software: (1) a reader controller in Java, (2) a data management wizard, and (3) a live visualization. The reader controller manages the hardware and the tracking system. The wizard is used to manage the database, which is most useful to configure events. Finally, the live visualization has been created within a web page using Processing.js. The visualization is procedurally generated out of the data retrieved from the web service and aims at improving social interactions by proposing a live graphical representation of geographic interests within the room.

A live test has been carried out with 120 participants during a forum-like event. The data is analysed and presented with graphical representations such as stacked bar charts, stream graphs, etc. From observations performed during the live event, a set of hypotheses on the behaviour of the participants has been established. The system supported all these hypotheses and confirmed the observations. Further extrapolations based on the data collected are presented to highlight participants’ behaviours that only the system can accurately track.

This Master Thesis has been achieved within a six-month time frame of combined research and implementation using agile methodologies (mainly Feature-Driven Development). Although work being performed by a single developer mitigates the value of agile methodologies, it allowed us to quickly iterate on the first transversal prototype.