A Narrow-Down Approach Based on H3 Cells for Indoor Localization

The popularity of the Wi-Fi fingerprinting approach is due to its cost-effectiveness and accessibility. However, it is vulnerable to long and short-term environmental dynamics and device heterogeneity. The high online computational complexity also causes delays in finding the user’s current location. To address these issues, we formulate a novel H3 cell-based two-stage, coarse and accurate positioning, narrow-down approach (NDA) that extracts and explores the useful features of access point (AP) measurements within each H3 cell. The coarse positioning algorithm comprises the clustering of Wi-Fi fingerprints into H3-cells, extraction of the useful features, and the estimation of the user’s coarse location using the Mahalanobis distance-based weighted K nearest neighbors (WKNN) algorithm while exploiting the knowledge of the test point (TP) fingerprint. The cell selection algorithms reduce the effects of environmental dynamics, device heterogeneity, and high online computational complexity. In the fine positioning phase, a set of support vector regressors (SVRs), trained on different sub-areas independently, predict the user’s final location within an acceptable time. The experimental results show that the proposed approach is robust to environmental dynamics and improves positioning accuracy significantly compared to the best solutions presented in the literature.