Landmark-Based Online Drift Compensation Algorithm for Inertial Pedestrian Navigation

Abstract

Pedestrian inertial navigation systems are nowadays integrated in all kind of location-based services. The orientation, particularly the heading estimation, is one of the most critical parts of the navigation system. The heading angle suffers from drift error. The algorithms proposed in the literature apply position corrections, which do not prevent the drift error to grow. We propose a novel landmark-based drift compensation algorithm, which allows the navigation system to generate drift compensated trajectories without the need of a post-processing stage. The chosen landmarks are corners and stairs, which are plentiful in indoor environments. Landmarks are seamlessly detected while the user walks. We test our algorithm with quasi-error-free turn rate measurements with known errors and also with medium-cost sensors. We provide in this work the proof of concept that the proposed algorithm is able to compute the underlying drift error and feed it back to the orientation estimator to generate drift compensated trajectories. Our algorithm is able to adapt to low drift changes during the walk, thus suited for long duration walks