AP location estimation method using crowdsourced Wi-Fi fingerprints

Abstract

An AP location is a valuable knowledge for Wi-Fi positioning, but surveying AP location is a demanding task. In this paper, we propose an AP location estimation method using implicitly crowdsourced data from the user's smartphone to reduce the cost while targeting an extensive area. The crowdsourced data is pairs of a Wi-Fi fingerprint and a location label from the location provider. The proposed method estimates AP location by optimizing the modified LDPL model to fit the data by applying weighted nonlinear least squares. Modifying LDPL model to reflect general environmental property and scoring influences of data increased the accuracy of the proposed method. Setting bounds reflecting the physical property for parameters of the model and filtering outlier data stabilized the proposed method. To evaluate the proposed method, the experiment was conducted on three datasets: KAIST N1, Cheongju airport, and Lotte World mall.