The availability of maps is one of the major prerequisites for deploying location-based services. However, only very few maps are publicly available for indoor environments. To overcome this problem, first approaches emerged to automatically derive indoor maps from odometry traces that pedestrian collected voluntarily. Although these approaches showed to be effective to automatically create indoor maps from mobility traces, they require energy-intensive positioning based on inertial navigation systems (INS) to collect traces of high quality, which impacts the users' willingness to participate in indoor mapping with his energy-constrained mobile device.

In this paper, we tackle this problem by providing an extension to automatic indoor mapping systems that lowers the energy consumption of the participating devices significantly. More precisely, we provide a framework enabling mobile devices to turn off INS while moving in areas that have been mapped already with high quality. In order to enable the dynamic re-start of INS---which requires an initial position and direction---when entering insufficiently mapped areas, we combine INS with low-energy WiFi-based position recovery. WiFi-based position recovery enables a coarse-grained position tracking while the inertial positioning is off and allows for bootstrapping INS by providing an initial position when needed.

Using our approach, we show that indoor models can be derived saving up to 25% of energy on the mobile devices without compromising on the quality of the derived maps.