Pedestrian localization with PDR supplemented by GNSS

Pedestrian localization in urban areas is often inaccurate, because GNSS signals could be blocked, attenuated or reflected by existing obstacles. The use of a combination of GNSS could reduce the drawbacks related to the scenario. Anyway, for scenarios particularly hostile, the multi-constellation approach is insufficient and the integration of multiple sensors is a possible solution. PDR is a technique based on inertial sensors, which exploits the characteristics of human gait. In PDR, the measurements from inertial sensors on a pedestrian are processed, with the aim of detecting the presence of steps, estimating their length and updating the pedestrian heading. In this paper, a PDR algorithm, processing measurements from accelerometers and gyros embedded in a smartphone, is implemented; a pedestrian carries the smartphone in texting pose. To reduce the PDR inherent drift, a loosely coupled integration architecture with GNSS is carried out; the data-fusion core is an extended Kalman filter, with position and yaw errors as inputs. The GNSS-based yaw is derived from velocity, which is usually obtained from Doppler measurements, with an accuracy, in good visibility condition, of cm/s order. In this work, a velocity derived from TDCP, with accuracy in the same condition of mm/s order, is used too. The yaw derived from TDCP allows significant improvements of PDR/GNSS performance. The adopted GNSS systems are GPS and Glonass, and a RAIM technique is applied to pseudorange, Doppler and carrier-phase measurements, in order to reduce the effect of outliers, which are very frequent in urban scenario.