Optimal Heading Estimation Based Multidimensional Particle Filter for Pedestrian Indoor Positioning

Abstract

Localization capability is a challenging task in global navigation satellite system-degraded or denied environments. Alternatively, today’s smartphones have an increased number of integrated sensors that can act as terminals for indoor personal positioning solutions such as pedestrian dead reckoning (PDR). However, magnetic interference, poor sensor measurements, and diverse handling of smartphone quickly decrease the performance for indoor PDR. This paper proposes a comprehensive and novel pedestrian indoor positioning solution in which heading estimation is improved by using simplified magnetometer calibration, by calculating projected acceleration along the moving direction using frequency-domain features and by applying direction constrains to indoor accessible paths. Moreover, compared with an ordinary particle filter (OPF) and a Kalman filter, this paper proposes a multidimensional particle filter (MPF) algorithm, namely MPF, which includes high-dimensional variables such as position, heading, step length parameters, motion label, lifetime, number of current particles, and factor. An MPF can handle more uncertain parameters than the OPF. Therefore, positioning with an MPF can achieve lower errors using low-quality sensors, mitigate interference introduced from surrounding environments, and reduce heading ambiguities due to different modes of carrying a smartphone. Consequently, field tests show that the proposed algorithm obtains robust performance for heading estimation and positioning.