A Novel Method Based on Particle Flow Filters for Stellar Gyroscope Parameter Estimations

Inertial sensors are commonly used in many applications for a while. They provide important information about the orientation but are prone to structural errors, which makes it impossible to use their raw data. Particularly for gyroscopes, deterministic errors such as bias, misalignment, or scale factor errors can affect their performance dramatically rather than stochastic ones. Thus, to determine and calibrate the parametric errors of the spacecraft gyroscope, sensor fusion methodology that ensures the integration with others such as star-tracker is needed alongside stochastic filtering. With robust filter structures, it is possible to get high performance even for unreliable low-cost sensors. So far, many filters are employed for gyroscope-error-parameter-estimations except the particle flow filter, which is an efficient non-linear filtering, especially for high-dimensional systems‘ state estimation. However, “the particle flow filter structure” is used for the prediction and calibration of gyroscope error parameters for the first time in the literature in this study. The “Flow of Particles” provides a novel approach to solving the curse of dimensionality of the standard particle filter for state estimations in a non-linear gyroscope error system here. Its performance differences with previous filters on parametric error estimations and their calibration are also highlighted. Although the particle flow filter-based algorithm has some computational complexity, it outperforms Kalman and particle filter-based approaches regarding accuracy for gyroscope parametric error estimations. Significant improvements in estimation errors are elaborated depending on filter factors such as particle numbers, flow step size, etc.