A collaborative approach to heading estimation for smartphone-based PDR indoor localisation

摘要: Pedestrian dead reckoning (PDR) is widely used for indoor localisation. Its principle to recursively update the location of pedestrian by using step length and heading. A common method estimate heading in PDR use magnetometer measurements. However, unlike outdoor environments, Earth's magnetic field strongly perturbed inside buildings making measurements unreliable estimation. This paper presents a new reduce estimation errors when magnetometers are used. The consists two components. first component uses machine learning algorithm detect whether within specific error margin. Only estimates margin retained passed second component, while other discarded. data fusion average from multiple people walking same direction. rationale this based on observation that perturbations often highly localised space if direction, then only some their likely be perturbed. Data between users can carried out distributed manner consensus with information sharing over wireless links. We tested performance our 92 datasets. shown provide an approximately 2°, which more than 6-fold lower raw (without any filtering fusion). Assuming accurate step-length observation, improved leads localisation accuracy 55cm, 80% improvement