Sensor fusion and rotational motion reconstruction via nonlinear state-observers

Despite the ease to instrument and record body accelerations and angular velocities of a body moving in space, the reconstruction from these measurements of rotational displacement is not a trivial task. Given that these quantities need to be integrated in order to define a displacement quantity the noise present in those signal can significantly disturb the results and limit the deployment of the technique in industrial applications, moreover the nonlinear kinematics that constrains the problem can be a challenge to the commonly used noise filtering techniques, such as linear state-estimators (Kalman filter). This paper aims to elaborate on the topic, by providing a concise formulation to the problem under rigid-body body assumptions and explore the use of nonlinear state-estimators to address the conditioning of the measured data, data fusion and reconstruction of the body motion. A comparison is drawn between an extended linear approach (EKF) and the proposed methodology, paying particular attention to the conditions that affect the performance of both methodologies. The paper compares results from numerical experiments using to better illustrate the differences between methodologies.