Performance Evaluation of a Stellar-Inertial Attitude Sensor Based on Particle Filtering

In this paper an attitude estimation approach relying on a particle filter is proposed for an autonomous attitude sensor which integrates rate gyros with a star sensor. Thanks to the complementary characteristics of modern inertial and star sensors, with the former being able of performing measurements at high data rates with precision decreasing over time, and the latter being able of high-precision measurements at lower updating frequency, the integrated sensor can provide complete operational autonomy during all the mission phases. Particle filters are Sequential Monte Carlo methods relying on a point-mass representation of probability density distributions, which allows generalizing traditional Kalman filtering approaches. Specifically, for inherently non-linear estimation problems, the particle filter allows overcoming the well-known limitations of the Kalman Filtering approaches, leading also to faster convergence from inaccurate initial conditions. The adopted filter formulation incorporates standard gyro-based model, for attitude propagation, and star sensor measurement model, for measurement update. The attitude parameterization is given in terms of Euler’s angles. The filter algorithm is implemented into a numerical model of the spacecraft orbital and attitude dynamics. Filter performance is numerically tested considering different dynamics conditions, and initial estimation errors. Simulation results show the effectiveness of the proposed filtering scheme in comparison with a standard Kalman Filter approach. They also indicate that the particle filter is more robust under realistic initial attitude-error conditions