Assessment of global orientation estimation using redundant inertial and magnetic sensors

Abstract

Inertial Measurements Units based on microelectromechanical technologies are extensively used in a vast range of applications due mainly to their reduced size, weight, power consumption and cost. However, the accuracy of the systems that rely on this technology is still the main hampering factor to its broad adoption. When it comes to global orientation estimation, the number of factors that play an important role in the final accuracy rise. It is possible to identify five important factors: sensor bias, sensor noise, algorithm robustness and tuning, and for the last, system bandwidth. Therefore this M.Sc. Dissertation details the development of a system with sensor redundancy, aiming to study the relation between inertial and magnetic data prevenient from different sensors, and finally, resulting in an improvement of the orientation estimation accuracy. Five different fusion algorithms were adapted and implemented, having also their parameters properly optimized. Additionally, a two degrees of freedom orientation system is developed and implemented to be used as orientation reference. The performed tests were also camera recorded for ease of visualization, the videos are available online. The results are presented in terms of Kinematic Tracking Error and angle Root Mean Square Error (RMSE). The proposed system was able to reach orientation errors as low as 1.05° RMS for the yaw axis during static tests (the yaw axis usually presents the highest RMS error), while dynamic experiments showed results as low as 1.17°, 2.30° and 3.03° RMS, respectively, for what has been classified as Slow, Mid and Fast movement profile. Furthermore, it is presented the hardware project of a semi-commercial system, which would be able to overcome the main issues presented by the actual system, what would be able to promote new researches.