Research On Spacecraft Attitude Estimation Method Based On Nonlinear Filtering Algorithm

The attitude estimation system based on the combination of gyro and star sensor is widely used in spacecraft with high accuracy requirements for attitude estimation.Since the spacecraft attitude estimation model exhibits strong nonlinear characteristics,the attitude estimation generally uses a nonlinear filtering algorithm.These algorithms are often an extended form of the Kalman filter.The nonlinear Kalman filter algorithm based on the Gaussian filter framework obtains the filtering solution on the premise that both the system noise and the measurement noise are Gaussian white noise.However,from the derivation equation of the dynamic model of the spacecraft or the discrete orbital spacecraft dynamics equation,the noise of the attitude estimation system is not Gaussian.And due to the influence of the attitude sensor,there is uncertainty in the measurement.Based on this consideration,at the same time,this is also an urgent problem to be solved in practical applications.In this paper,the adaptive filtering algorithm,robust filtering algorithm and particle filtering algorithm for spacecraft attitude estimation are studied.For the attitude information output of the star sensor in the attitude estimation system,there are uncertainties such as measurement loss.In this paper,a modified unscented quaternion attitude estimation algorithm is derived.The attitude estimation model with different loss rates is established.The filtering process based on quaternion and three parameters is given.The stochastic stability of the designed algorithm is verified theoretically based on Lyapunov function.At the same time,the boundedness of the error covariance is analyzed.From the stability analysis,the influence of the loss rate on the filter stability is found,and a scheme for adding adjustment factors is proposed based on the derived mathematical inequality.In view of the two types of simulation scenarios of the presence or absence of measurement loss,two adjustment schemes and adjustment factors are designed.In the simulation experiment,the small-angle initial condition error and the large angle initial condition error are tested,and different loss rates are considered.The experimental results show that the designed algorithm has better estimation performance.For the attitude estimation model with dynamic or measurement model error and non-Gaussian noise,an adaptive robust cubature Kalman filter algorithm based on an optimal adaptive factor,an adaptive robust cubature Kalman filter algorithm based on a multi-fading factor,and nonlinear attitude estimation algorithms based on additive quaternions are proposed in this paper.Among them,the optimal adaptive factor is derived based on the estimated covariance matrix of the prediction residual.The multi-fading factor is based on the strong tracking algorithm and considers the measurement vector of the attitude model for derivation design.The filtering algorithm based on the additive quaternion adopts Huber-based robust filtering method and the adaptive filtering method based on random fading factor.In the simulation experiment,the adaptive robust algorithms based on quaternion and three-parameter switching are compared with the traditional attitude estimation algorithms and the adaptive robust algorithm in the literature in five cases.The proposed algorithms have obvious advantages and have higher estimation accuracy and good robustness.The filtering algorithms based on additive quaternion are also compared in several sets of experiments,and the selection of filtering parameters is also analyzed experimentally.Aiming at the problem of particle degradation and particle deficiencies of particle filtering in attitude estimation system,this paper proposes an adaptive genetic particle filtering algorithm suitable for attitude estimation system.Gaussian filtering algorithm is used to generate the importance proposal distribution in the particle filter framework.Before resampling,the crossover in the genetic algorithm step and the improved mutation strategy are used to improve the particle diversity.The proposed adaptive strategy can automatically determine the mutation probability of the designed algorithm.The simulation experiments show that the proposed adaptive genetic particle filtering algorithm is outstanding in the univariate growth model and the comprehensive experiment and has better filtering performance than the benchmark algorithm and other particle filtering algorithms.After regularization of adaptive genetic particle filtering algorithm,the designed algorithm outperforms other attitude estimation algorithms under large initial conditional error conditions.