Experimental Comparison of Sensor Fusion Algorithms for Attitude Estimation

Abstract

Inertial Measurement Unit is commonly used in various applications especially as a low-cost system for localization and attitude estimation. Some applications are: real-time motion capture system, gait analysis for rehabilitation purposes, biomedical applications, advanced robotic applications such as mobile robot localization and Unmanned Aerial Vehicles (UAV) attitude estimation. In all the mentioned applications the accuracy and the fast response are the most important requirements, thus the research is focused on the design and the implementation of highly accurate hardware systems and fast sensor data fusion algorithms, named Attitude and Heading Reference System (AHRS), aimed at estimating the orientation of a rigid body with respect to a reference frame. A large number of different solutions can be found in the literature, and an experimental comparison of the most popular is presented in this work. In particular, the algorithm based on the gradient descent method and the algorithm based on a nonlinear complementary filter are compared to a standard Extended Kalman Filter (EKF) with the aim to show that a general method can easily compete with ad-hoc solutions and even outperform them in particular conditions. In order to validate the estimation accuracy a Kuka robot is used to compute the ground truth. Moreover, in order to estimate the computational burden, the algorithms are implemented on an ARM-Cortex M4-based evaluation board.