Hardware-Independent Automatic Crowdsourcing-Based Hybrid WLAN-RFID Adaptive Indoor Tracking System Using Fast Orthogonal Search and Multiple Particle Filters

摘要

This paper introduces a novel methodology to improve indoor tracking systems in local area wireless networks(WLAN) based on received signal strength (RSS). The proposed methodology addresses significant shortcomings and drawbacks in current existing indoor WLAN-based tracking systems. First, it does not need offline calibration or manual data collection. Instead, it uses an automatic crowdsourcing-based technique with a Fast Orthogonal Search algorithm to process sparse unequally-spaced RSS measurements. Second, it solves the hardware variations problem where multiple mobile devices from different manufacturers with different RSS levels are to be tracked. Third, the proposed system handles both short-term and long-term RSS variations through a novel multiple-particle filter approach. Finally, tracking is performed through a tightly-coupled particle filter algorithm that fuses RSS observations with a simple pedestrian walking model. Experimental and simulation work showed that the proposed system can efficiently build RSS models for different user's devices hardware, greatly smooth RSS measurements and track users' devices with an average of 2m accuracy.