Adaptive WiFi positioning system with unsupervised map construction

A major problem in indoor localization based on WiFi RSS (Received Signal Strength) is the tedious offline surveying process for fingerprinting. To address this problem, we propose an unsupervised Simultaneous Localization and Mapping (SLAM) system for automatic floor map and radio map construction. All it takes to set up this system is for a surveyor to walk through the coverage area randomly several times to collect traces of WiFi RSS measurements. Based on similarity matching of these measurements, a floor map in form of a graph is automatically constructed, together with a radio map which associates individual WiFi RSS distributions with sample points along the graph. In the online stage, this system can be dynamically updated using crowdsourced RSS data. Algorithms are also discussed to increase the computational efficiency, as the size of the database can increase squarely with the scale of the test bed.