InPReSS: INdoor Plan REconstruction Using the Smartphone's Five Senses

Today, we can use our smartphones to drive from Los Angeles to New York, yet we can not use them to find our way around Charles de Gaulle Airport. GPS simply does not work indoors and while outside areas are mapped very well, the indoors of most buildings remain a mystery to our favorite navigation apps. While the problem of indoor localization has attracted a lot of attention, with several solutions with meter-level accuracy emerging as research prototypes, the problem of building accurate and detailed indoor maps has been neglected. We present InPReSS, a solution for dynamic indoor map construction without explicit user input. It leverages the sensors available on off-the-shelf smartphones and their ubiquity to construct floor plans as users go about their daily business. The InPReSS approach consists of four steps: 1. Collecting readings from five sensors, tagged with their location from smartphones available in the target building, and divide the floor plan into cells using a Voronoi diagram. 2. Clustering the cells into rooms based on the sensor readings. 3. Identifying doors and ways between adjacent rooms using the Voronoi diagram and user movement traces. 4. Building a two-level graph of the floor plan for navigation.