Characterization of the Impact of Indoor Doppler Errors on Pedestrian Dead Reckoning

Indoor pedestrian navigation is a very challenging task because self-contained sensors are affected by instrumentation errors corrupting the navigation solution and GNSS signals, which could be used for calibrating the latter, are barely available. Doppler measurements are found to be more accurate than pseudoranges in these harsh indoor environments, especially with narrowband receivers. Therefore the impact of indoor Doppler errors on a Pedestrian Dead Reckoning (PDR) navigation filter is investigated. Doppler errors are simulated using experimental data post-processed with the high sensitivity GSNRx-ss (TM) software receiver and a derived Doppler error model. Step length and heading errors are simulated for a 500m pedestrian walk. All these controlled errors are introduced in a novel tight PDR/Doppler coupling Extended Kalman filter for assessing the impact of Doppler indoor errors on the navigation solution. It is found that even biased Doppler measurements control the errors' growth in the navigation solution, principally in the attitude angles estimates but also in the norm of the velocity vector. With a 150 error in the walking direction, the horizontal position error equals 10% of the travelled distance for the coupled PDRIDoppler solution and 20% for the MEMS only solution. The analysis highlights also the need for designing new methods to discard outliers in the set of indoor Doppler measurements and benefit from new indoor GNSS observations.