Adaptive Fingerprint Database Update Method Based on Gaussian Process Regression for Indoor Localization

With the popularity of application services based on indoor location recently, Wi-Fi indoor location based on signal fingerprints has attracted wide attention. The location accuracy of this method depends on the matching degree of current perceived signal strength information and fingerprint database. However, Wi-Fi signals are varied dynamically with changes in the external environment. In this paper, an automatic fingerprint updating scheme based on heading features with prior knowledge (AfuBpk) is proposed to mitigate the negative impact of access point (AP) change and dynamic environment on localization results. This scheme updates fingerprint database automatically without additional offline acquisition process. The collected Wi-Fi signal and the heading features extracted from the inertial navigation sensors are used to judge whether AP changes. When AP changes are detected, the current fingerprint database is calibrated and updated using Gaussian Process Regression (GPR). Experimental results show that the proposed method has a more stable localization performance compared with the existing methods when the signal strength changes due to the external environment.