A novel motion-reconstruction method for inertial sensors with constraints

Abstract

Motion reconstruction for rigid bodies and rigid-body frames using data from inertial measurement units (IMUs) is a challenging task. Position and orientation determination by means of IMUs is erroneous, as deterministic and stochastic errors accumulate over time. The former of which errors can be minimized by standard calibration approaches, however, sensor calibration with respect to a common reference coordinate system to correct misalignment, has not been fully addressed yet. The latter stochastic errors are mostly reduced using sensor fusion. In this paper, we present a novel motion-reconstruction method utilizing optimization to correct measured IMU data by means of correction polynomials to minimize the deviation of motion constraints. In addition, we perform gyrometer and accelerometer calibration with an industrial manipulator to address deterministic IMU errors, especially misalignment. To evaluate the performance of the novel methods, two types of experiments, one at constant orientation and another with simultaneous translation and rotation, were conducted utilizing the manipulator. The experiments were repeated for five individual IMUs successively. Application of the calibration and optimization methods yielded an average decrease of 95% in the maximum position error compared to the results of common motion reconstruction. Moreover, the average position error over the measurement duration decreased by nearly 90%. The proposed method is applicable to velocity, position, and orientation constraints for every experiment that starts and ends at standstill.

Accurate Estimation of Robotic Arm Movements for Effective Motion Control: Utilizing Multiple Sensors and Data Fusion

“…The estimated joint angles consistently fell within therapeutic limits. Neurauter and Gerstmayr (2023) published an article that specifically addresses motion reconstruction challenges using IMUs for rigid bodies and proposes a novel method that incorporates optimization and correction polynomials to minimize motion deviation. Experimental results with an industrial manipulator demonstrate substantial reductions in position errors, achieving a 95% decrease in maximum error and an average reduction of nearly 90% throughout the measurement duration.…”

Particle tracking in snow avalanches with in situ calibrated inertial measurement units

“…In principle, IMUs are designed to enable the determination of angular orientation, translational velocity and position. While the reliability of angular orientation is usually high, the one of velocity and, even more, of position suffer from an unavoidable drift occurring in the course of an unavoidable integration procedure (Neurauter and Gerstmayr, 2022). To obtain long-term stable results, additional information is required (GNSS data, e.g.…”

Particle trajectories, velocities, accelerations and rotation rates in snow avalanches

“…Orientation with respect to the earth frame can be derived via a Gram-Schmidt projection utilising standstill data from the accelerometer and magnetometer (Neurauter and others, 2021). Due to drifting accelerometer data, special motion reconstruction algorithms are applied that require specific IMU data processing and a corresponding sensor calibration (Neurauter and Gerstmayr, 2022).…”

“…The whole dataset from this measurement is multiplied with this factor. This does not include an in depth calibration which would consider an sensor offset and corresponding time drift (Neurauter and Gerstmayr, 2022). The accelerations and rotation rates are not smoothed at all, because when heavy impulsive impacts occur during the experiment, it should be noticeable in the measurement data.…”

“…Thus we conclude, that inertial navigation by IMU data is limited and requires additional post-processing steps to handle errors, especially acceleration drift. Despite errors in accelerations, rotation rates derived from measured angular velocities were accurate and yielded meaningful information regarding particle rotations (Neurauter and Gerstmayr, 2022). To take advantage of the redundancy in translational movement and the high sampling rate of IMU data, future work will be conducted in the field of sensor fusion.…”