AGV-Assisted Construction of Dynamic Wi-Fi Fingerprint Databases for Indoor Localization

Wi-Fi fingerprinting has become a dominant technology for indoor localization because it does not require additional hardware or new infrastructure. However, the construction of fingerprint databases remains one of its major challenges because collecting a large number of fingerprints over service areas is not only labor-intensive but also time-consuming. The high labor and time cost becomes even more challenging for the construction of dynamic fingerprint databases to deal with the time-varying nature of wireless channels indoors, where Wi-Fi Received Signal Strength Indicators (RSSIs) are to be measured at the same Reference Points (RPs) repeatedly. To address these issues, Automated Guided Vehicle (AGV)-assisted construction of dynamic fingerprint databases is proposed in this paper. As it is assumed that users use their smartphones for indoor localization service, a smartphone is mounted on an AGV and used to collect Wi-Fi RSSIs every 10s as the AGV moves along a pre-arranged path with reference points marked with tag36h tag family for AprilTag 3. Then, the locations of the collected RSSIs are aligned to those of the tags based on the time information of both AGV and smartphone. The analyses based on the AGV-collected data on the three floors of the XJTLU IR building and the corresponding data from the XJTLU dynamic database, the latter of which are jointly collected by human participants and wallmounted Raspberry Pi Pico Ws, demonstrate that the proposed AGV-assisted construction of dynamic fingerprint databases can be used in complex indoor environments and reduce the high labor and time cost of the manual data collection without impairing the indoor localization performance.