Orientation Power - LowOn the Accuracy Improvement of 1 IMU and MARG Sensor Arrays Usings Filter

The orientation estimation filter proposed by Madgwick [1] for inertial and inertial/magnetic sensors have been successfully used as a core solution in a variety of commercial low-power motion tracking devices. The approach offers a high accuracy and reduces the computational cost compared to the state-of-the-art Kalman-based methods [10]. In this paper, we propose a modification to the Gradient Descent correction step in [1], which guarantees the improvement of accuracy in quaternion estimation without increasing the total number of required arithmetic operations. Monte Carlo simulations with typical statistical characteristics for sensor errors have shown an average 37.2% improvement in dynamic root-mean-square error of Euler angles.