Efficient Sequence Matching and Path Construction for Geomagnetic Indoor Localization

Geomagnetic field is highly applicable to indoor localization due to its pervasive spatial presence, high signal stability and infrastructureless support. Previous work in the area often fuses it with pedometer (step counter) via particles. These approaches are computationally intensive and require strong assumptions on user behavior. To overcome the weaknesses, we propose Magil, a pedometer-free approach leveraging solely upon magnetic field for indoor localization. To the best of our knowledge, this is the first piece of work using only geomagnetism for smartphone localization. Magil is applicable to any open or complex indoor environment. In the offline phase, it continuously collects and stores geomagnetic fingerprints while a surveyor is walking in pre-defined paths covering the indoor area. In the online phase, it identifies the indoor segments whose fingerprint variations highly match with the target observations. With a modified shortest path formulation, Magil selects and connects these matched segments and obtains the target locations and paths. We have implemented Magil, and conducted extensive experiments in our university campus. Our results show that Magil outperforms many state-of-theart schemes by a wide margin (cutting localization error by more than 30%).