Kernel density-based radio map optimization using human trajectory for indoor localization

Accurate indoor localization remains a significant challenge due to the complex nature of indoor environments. This paper proposes a novel method for constructing a radio map (RM) based on Kernel density estimation (KDE) and human trajectories (HT) to enhance indoor localization accuracy. The proposed method utilizes historical HT data in RM construction to capture the spatial variability and complexity of indoor environments, which is crucial for accurate localization. By employing KDE, kernel density maps are generated, identifying high-density regions where additional interpolated fingerprints are strategically placed to improve localization accuracy. In contrast to the conventional method of uniformly placing interpolated points (IPs), the proposed approach better models natural walking patterns and trajectories, thereby enhancing the uniqueness and accuracy of user position identification. Through extensive experiments with various HT patterns, the proposed KDE-RM optimization method consistently outperforms the conventional approach of evenly distributed IPs using Kriging and inverse distance weighting interpolation by up to 36.4%. This demonstrates the effectiveness and potential of the proposed method as a valuable tool for enhancing indoor localization.