Easing the survey burden: Quantitative assessment of low-cost signal surveys for indoor positioning

Indoor positioning through signal fingerprinting is a popular choice since it requires little or no additional infrastructure. However, the initial creation and subsequent maintenance of the signal maps remains a challenge since traditional manual surveying is not scalable. In this work we look at the use of path surveys, which attempt to construct the signal maps from a sparse set of fingerprints collected while a person walks through a space. As such, the survey points rarely provide a uniform coverage of the space of interest. We quantitatively evaluate path surveys with reference to a detailed manual survey using smartphone-grade equipment. We compare both the individual maps (generated using Gaussian Process regression) and their collective positioning performance. Our results are for both WiFi and Bluetooth Low Energy signals. We show that a path survey can provide maps of equivalent quality to a manual survey if a series of guidelines we provide are followed.