Evaluation of Inertial Sensor Fusion Algorithms in Grasping Tasks Using Real Input Data: Comparison of Computational Costs and Root Mean Square Error

Comparison of a sensor fusion algorithm implementation on a C674X DSP and a Cortex A8 core

“…Results in [16] show, that this Kalman filter shows significantly less computational time and a higher accuracy compared to several other Kalman filters in literature. Table 1 lists the number of basic floating point arithmetic operations required for performing a single filter cycle, for a single inertial sensor and sampling instance.…”

“…For magnitudes outside this range, the inputs are computed based on previous step orientation estimation and reference vectors. A previous study showed that this filter achieves significantly better results and demands less computational load compared to several other inertial sensor fusion approaches [16].…”

Development of a wireless low-power multi-sensor network for motion tracking applications

“…Common classes of algorithms estimating the orientation are based on three steps: integration, vector observation and Kalman filtering [6]. The integration step predicts the state by means of the gyroscope data, the vector observation step provides the optimized observation using the magnetometer and the accelerometer raw data, and finally, the Kalman filter is applied, providing the optimal estimators which minimize the error covariance.…”

An inertial human upper limb motion tracking method for robot programming by demonstration

“…Pure IMU orientation usually catches up the gyroscope drift only along direction perpendicular to gravity. Several methods have been explored [50,51]: complementary filters [52,53,54]; Kalman and Extended Kalman filters [55,56,57], gradient descent filters [58] or integration and vector observation [59,60].…”