Assessing a 6C Kalman Filter using Experimental Datasets from an Industrial Robot

Abstract:

Shaking from earthquakes is a complex set of six components (6C) that make up the ground motion, containing both translations and rotations. Nonetheless, state-of-the-art earthquake monitoring stations only include accelerometers and GNSS to record the strong ground motion. The inertial accelerometer sensors directly measure the translational part of the motion, though rotations contaminate this output and are difficult to quantify. GNSS is less sensitive to rotations but can only resolve large amplitudes and long period ground motion. Therefore, the future design of monitoring stations should include a rotational sensor to allow a complete reconstruction of the ground motion.

We have built a monitoring station that combines an accelerometer and rotational sensor both sampling at 250 Hz, with a GNSS receiver and antenna determining coordinates with a lower sampling rate of 100 Hz. The three instruments have been fixed to a platform attached to the top of an industrial six-axis robot arm, while the arm performs simulated ground motions with high accuracy and repeatability. The robot records its own feedback loop including position and orientation of the platform, which serves as the ground truth of the performed trajectory. This allows us to compare all the results obtained to the actually performed trajectory of the robot.

We designed a 6C Kalman filter that combines the three instrument records using an optimal equation design and optimal tuning of the parameters. It includes time domain tilt correction of the acceleration and subsequent estimation and correction of the instrument records to get the state vector consisting of displacement, velocity and rotation. The developed framework and sensing scheme is able to output rotation-free, broadband and precise 3C translations as well as drift-free and precise 3C rotations.