Robust and fast algorithm design for efficient Wi-Fi fingerprinting based indoor positioning systems

Abstract

Indoor positioning systems (IPS) based on Wi-Fi fingerprinting have gained significant attention due to their potential for providing location-based services. Large scale IPS deployments require implementation of robust, accurate, and fast algorithms. Data analytics assisted algorithms provide positioning accuracy improvements, however their integration within real-time and scalable solutions significantly depends on the computational complexity. Therefore, we propose robust and computationally efficient algorithms for performance enhancements in Wi-Fi fingerprinting-based IPS. A robust radio map algorithm based on enhanced statistical cluster initialization was designed for efficient indoor environment characterization. The proposed data filtering algorithm leveraged smart clustering to mitigate real-time data variations. The designed area classification algorithm was based on smart dual-band data aggregation. We evaluated the performance of proposed algorithms based on accuracy and computation time. The performance evaluations signified accuracy and computational enhancement, in comparison to related benchmark techniques. The data filtering and area classification algorithms required 40% less computation time. Simultaneously, 14.5% and 36% accuracy improvements were recorded for the area classification and radio map algorithms respectively. The proposed algorithms have the potential to significantly enhance IPS performance in a variety of real-time applications, including indoor navigation and asset tracking.