Adaptive Integrated Indoor Pedestrian Tracking System Using MEMS sensors and Hybrid WiFi/Bluetooth-Beacons With Optimized Grid-based Bayesian Filtering Algorithm

摘要

With recent dramatic increase in sensors deployments and processing nodes, accurate indoor positioning, tracking, and navigation is becoming achievable. Among many platforms that need to be localized and tracked are pedestrians. A reliable indoor pedestrians tracking has a wide range of applications such as healthcare, retail, rescue missions and context-awareness applications. This paper introduces a calibration-free hybrid indoor positioning system that utilizes inertial sensors (INS), wireless local area networks (WLAN), and low-cost Bluetooth low-energy (BLE) wireless beacons. BLE beacons are becoming very popular in retails and they can be easily installed in any indoor environment. To deal with jumpy and noisy nature of wireless positioning indoors, and to cancel out the unbounded drifts associated with INS, this work investigates the utilization of Grid-based nonlinear Bayesian filtering to fuse all the aforementioned sensors measurements. The motion-updated prior is modeled as a probability distribution (PD) that takes discrete values (cells) with a pre-defined resolution. This prior PD is shaped by the pedestrian dead-reckoning model (PDR). The measurements-updated posterior PD is calculated using a numerical approximation of Bayes' rule where measurements are signal strength observations. The measurement model is mainly a tightly-coupled model where all motion states, sensors errors, and wireless propagation models are represented. To further enhance the accuracy, all RSS measurements are filtered out by a separate Kalman filter with a Gauss-Markov process model. The performance is further enhanced by applying measurements updates from BLE beacons. The system is realized on an experimental setup that consists of off-the-shelf INS existing on smart-phone and WiFi. For flexibility and analysis purposes, the BLE beacons were simulated on top of the collected INS/WiFi measurements. The ground-truth was obtained by an unmanned ground vehicle (UGV) equipped by INS and Odometer aided by floor maps.