An Improved Kriging Algorithm for Spatial Variability in Wi-Fi Fingerprint Positioning Database Construction

In Wi-Fi fingerprinting location, constructing traditional fingerprint databases consumes a large amount of human resources. Interpolation algorithms are typically used to reduce the workload of database construction, and the Kriging algorithm is a widely used interpolation algorithm. However, existing algorithms do not consider the nonlinear variation in received signal strength indication (RSSI) in free space. In this paper, an improved Kriging algorithm based on spatial variability is proposed by analysing the propagation law of electromagnetic waves in free space, separating the regions with drastic and slow changes in RSSI, and then fitting a semivariance function to the two regions for interpolation. Through real simulation tests, the proposed algorithm has demonstrated improvements in interpolation accuracy, achieving gains of approximately 4.6 % and 8.1 % over traditional Kriging and the Inverse Distance Weighting (IDW) algorithm, respectively. The interpolation accuracy of the proposed algorithm is superior to both Kriging and IDW algorithms, effectively enhancing localization precision, making it more suitable for constructing an RSSI fingerprint database.

I. Introduction

With the rapid development of information technology and the popularization of various smart devices, Wi-Fi fingerprinting [1]–[2] has recently become a popular research topic [3]–[5]. Fingerprint-based localization is a Wi-Fi positioning method for Integrated Sensing and Communications(ISCC) and has a wide range of applications. However, in the offline phase [6]–[8], constructing a fingerprint database requires manual point- by-point collection, resulting in high labour costs. Moreover, data collection is usually restricted and unavailable in certain places such as offices or private houses. Therefore, a spatial interpolation algorithm is the most effective solution for re-ducing the human resources required for fingerprint database construction without affecting its accuracy.