Improved Wi-Fi AP position estimation using regression based approach

This paper describes the improved Wi-Fi AP position estimation method for building more accurate Wi-Fi AP position DB in complex indoor signal propagation environment. One of our pro Fi AP by using indoor survey for higher Wi-Fi AP position accuracy. Our contribution focuses on the Wi-Fi AP position estimation method. In previous works, there were several methods for Wi-Fi AP position estimation by using such as weighted centroid localization (WCL), the trilateration using traditional signal propagation model etc. In case of WCL, the estimated Wi-Fi AP position depends on the amount and location of collecting points (CP) so that the Wi-Fi AP position accuracy can be poor in low DoP (Delusion of Precision) collecting environment, like the corner or edge of a building. Trilateration method can be more efficient than weighted centroid in case of low DoP collecting scenario but its method using the traditional path loss model is vulnerable to complex indoor area where lots of attenuation, reflection and refraction exist. To enhance Wi-Fi AP position accuracy in case of low DoP collecting environment, the Wi-Fi AP position estimation using Gaussian process regression (GPR) is proposed. Firstly, the Wi-Fi AP received signal strength (RSS) using GPR is predicted and analyzed whether it can be used to estimation Wi-Fi AP position. Next, the comparison of Wi-Fi AP position accuracy by using WCL with by using GPR/WCL is described. From the test result, we will show that the position accuracy of the proposed method will be better than using WCL.