A survey of crowdsourcing-based indoor map learning methods using smartphones

Abstract

Location-based services and applications have grown rapidly over the past decade. These location-based services and applications usually use maps to display user locations. However, the availability of digital indoor maps and indoor route information is far from satisfactory. At present, most indoor applications rely on manually created indoor maps, which requires huge costs and effort for editing and maintenance. Because mobile phones have rich sensors, scientists hope to explore the use of smartphone crowdsourcing to generate indoor maps in recent years. This article provides a systematic review of these works. Unlike former surveys, we divide the indoor maps into indoor route maps and indoor floor plans. For these two kinds of indoor maps, we summarize their common steps and differences in the crowdsourcing-based map learning process. Basically, they share the trajectory collection, landmark detection, trajectory alignment, and graph optimization steps. The floor plan learning needs additional trajectory segmentation and boundary extraction steps. We introduce the state-of-the-art techniques of each of these key processing steps. Finally, we systematically compare existing map learning systems of these two categories in terms of sensor used, users’ participation types, and key technologies used in different steps.