Passive Crowdsourcing of Indoor Environment Features with Mobile Devices (Masterthesis)


Christian Spieß

12.11.2013, 16:00, room 4981


In the past two decades, mobile devices such as navigation systems, smartphones, or tablets have become affordable and powerful for many fields of application. Each of these devices are capable to position the user and direct him from a certain start to a certain end point. The continuous position estimation is done either by means of relative positioning, for example, inertial navigation systems, absolute positioning, for example, global satellite navigation systems, or through the fusion of both. For the purpose of indoor navigation, GNSS are mostly not available or very inaccurate. Moreover, in most cases a map for positioning based on environment features is not existent for indoor environments. Actively map these areas with only a small group of people is time consuming, expensive and slow in reaction to environment changes. This demands to develop methods that allow to passive crowdsource the mapping.  This thesis describes a method to obtain a feature map based on smartphone sensor data. To this end, the Wi-Fi graph-based SLAM approach, first presented in [46] is explained in detail.  Moreover, a trace matching algorithm is proposed to estimate a feature map of multiple traces and thus crowdsource the map estimation. In addition, an evaluation of the optimized traces and the matching is given, and the accuracy with respect to ground truth is compared to state-of-the-art indoor positioning systems