Towards Ubiquitous IPS: Leveraging Crowdsourced Data Accumulation over Time to Alleviate Reliance on External Sources in Initial Fingerprinting Map Generation

The rising demand for location-based services (LBS) underscores the critical need for accurate and ubiquitous indoor positioning systems, essential for supporting a wide array of applications across various industries. Over the past two decades, fingerprinting-based positioning methods utilizing pervasive WiFi received signal strength measurements have provided enhanced solutions. However, the manual creation and maintenance of fingerprinting databases are highly labor-intensive and significantly limit scalability. Crowdsourcing offers a scalable solution by engaging regular users in the creation of offline databases, thus eliminating the necessity for expert involvement. Despite this, challenges remain in the need for localization adjustments and calibration sources to ensure the generation of reliable offline databases. Most studies depend on external sources, such as floor plans, deployed anchor nodes, or feedback from active users, which can impede the development of a ubiquitous system. In this paper, we propose leveraging crowdsourced data accumulated over time to automatically infer the positions, and propagation characteristics of fixed WiFi access points to act as anchors with known locations to align and calibrate the crowdsourced traces. The inferred pervasive anchor nodes improved the fingerprints localization and expanded the radio map coverage.