Section: New Results

Experimental evaluation of attitude estimation algorithms for smartphones

• Context: Pervasive applications on smartphones increasingly rely on techniques for estimating attitude. Attitude is the orientation of the smartphone with respect to Earth’s local frame.

  Modern smartphones embed sensors such as accelerometer, gyroscope, and magnetometer which make it possible to leverage existing attitude estimation algorithms.

• Contribution: We investigated the precision of attitude estimation algorithms in the context of commodity smartphones carried by pedestrians. We considered eight typical motions (such as texting, phoning, running, etc.) with various impacts on external accelerations, as well as the presence/absence of magnetic perturbations typically found in indoor environments. We systematically analyzed, compared and evaluated eight state-of-the-art algorithms (and their variants). We precisely quantified the attitude estimation error obtained with each technique, owing to the use of a precise ground truth obtained with a motion capture system (the Inria Kinovis platform). We made our benchmark available (see Sec. 5.1 above) and payed attention to the reproducibility of results. We analyzed and discussed the obtained results and reported on lessons learned [7] [17]. We also presented a new technique which helps in improving precision by limiting the effect of magnetic perturbations with all considered algorithms.