Triggered INS/GNSS Data Fusion Algorithms for Enhanced Pedestrian Navigation System

Abstract

The paper deals with Personal Navigation Systems based on sensors commonly found in small devices such as smartphones and wearables. This research field has recently received broad interest within the scientific community thanks to many novel applications, which span from health and entertainment up to human navigation, rescue operations in dangerous environments, goods delivering, and others. The proposed approach features an outdoor navigation system obtained via an event-triggered multi-rate size-varying Kalman filter, which exploits the complementary properties of both a ground-based positioning system and an on-board navigation system, namely a Global Navigation Satellite System and a Pedestrian Dead Reckoning algorithm based only on inertial measurements. The data fusion method is designed to exploit real-time estimates of heading and step length, provided by the PDR, with the position obtained from GNSS. The presented solution can be used to recognize how the time-varying features of the user gait evolve, and to remove systematic errors due to wrong calibration of the sensors or to environmental noises. The proposed implementation can also work at each step with reduced information from the sensors, thus covering both the case of missing data (e.g., poor satellite coverage), and the situation of different sampling rates for each source. This latter scenario can be enforced to strongly decrease the use of GNSS and so to improve the efficiency for future implementations in low-energy devices. Experimental results show both the accuracy of the basic algorithm and possible applications of the extended features.