Unscented Kalman Filter-Based Fusion of GNSS, Accelerometer, and Rotation Sensors for Motion Tracking

In this paper, we present an unscented Kalman filter (UKF) for fusion of information from an accelerometer, global navigation satellite system (GNSS) instrumentation, and rotational sensor recordings of structural motion. Seismic and structural motions do not only include translations, but further incorporate torsion and twisting of the ground and/or structural components. Accelerometer and GNSS positions are known to be prone to errors introduced by rotation, such as (1) gravitational leakage, (2) misorientation, and (3) antenna pole tilt. In alleviating such effects, we propose fusion of information from six component (6C) data-3C translation and 3C rotation-and demonstrate its applicability for motion tracking on a flexible pedestrian bridge. To simulate a variety of load effects, the bridge was subjected to various sources of excitation such as hammer impulses, jumping, twisting, and running, as well as a combination thereof named the "artificial coupled forcing." The rotation errors of both the accelerometer and GNSS-estimated positions are corrected via a UKF-based fusion. We further identify the modal properties of the monitored bridge, excited by the different excitation sources, using a covariance driven stochastic subspace identification. The twisting of the bridge is shown to be a primary source of rotation errors. These errors ought to be corrected because their order of magnitude can be as large as the actual signal in the case of GNSS positions and up to 10% for accelerometer sensors. We compare the proposed UKF-based fusion for 6C motion tracking against a simplified linear Kalman filter and demonstrate the potential of the former for real-time, broadband, rotation-free displacement, velocity, and rotations tracking.