Elsevier

Information Fusion

Volume 18, July 2014, Pages 33-42
Information Fusion

A switch-mode information fusion filter based on ISRUKF for autonomous navigation of spacecraft

https://doi.org/10.1016/j.inffus.2013.04.012Get rights and content

Abstract

Aiming at the problem of loss of accuracy using extended Kalman filter (EKF) in case of orbit maneuver, this paper proposes a novel information fusion filtering algorithm-iterated square root unscented Kalman filter (ISRUKF), and then designs a switch-mode information fusion filter based on ISRUKF and extended Kalman filter (EKF). This method combines navigation sensors’ geocentric vector and geocentric distance with starlight angular distance, which efficiently improves the reliability of autonomous navigation. On this basis, the method deduced measurement function of information fusion. With a semi-physical simulation to verify the proposed method, the simulation results for stably running and orbital maneuvering spacecraft show that the switch-mode information fusion filter can reduce the complexity of the algorithm and ensure the accuracy of the estimation. Thus, the proposed switch-mode filter is very suitable for spacecraft autonomous navigation system and other strong nonlinear state estimation fields.

Introduction

Nowadays, the position of spacecraft is mostly determined based on the observation from ground stations. Therefore, there exists a tendency towards autonomous orbit determination of spacecraft, which can realize less cost on satellites and less ground-based operations. It is needed to note that the selection of measurements and filtering algorithm play determinative roles in navigation accuracy.

For the present there are some measurements for autonomous navigation of spacecraft, including geocentric vector and geocentric distance of earth sensor and star sensor’s outputs. Though this kind of navigation system shows good results, its accuracy depresses apparently in case of orbit maneuver. Meanwhile, sensor technology has wider and wider application fields such as autonomous navigation of spacecraft. Some scholars proposed integrated navigation to deal with orbit maneuver, such as inertial/celestial navigation system [1], which adds the complexity of previous system. In fact, starlight angular distance is an important measurement of navigation system, which includes stable orbit information of spacecraft and does not need additional navigation sensor. So we combine sensors’ geocentric vector and geocentric distance with starlight angular distance to realize autonomous navigation.

Meanwhile, EKF is the most widely used method in the field of autonomous navigation, which linearizes the process and measurement models up to first order term of Taylor series expansion. This linearization can lead to unavoidable large errors and may damage the convergence of the algorithm [2]. Scholars have also studies derivative-free alternatives such as UKF introduced by Julier and Uhlmann [3], [4], which approximates the probability density by evaluating the nonlinear function with sigma points. Rambabu et al. [5] applied the UKF and EKF for nonlinear state estimation. They compared the two methods by simulations after analyzing the principle of the two algorithms, and drew the conclusions that the UKF outperformed EKF in robustness and convergence rate. However, the UKF is unsatisfactory when the system is of weak observability and large initial error. For these reasons, some scholars proposed IUKF, which was developed under the general UKF framework and can adjust the state estimation to adaptively approach the true value through corrections of the measurement [6]. Meanwhile, Ref. [7] proposed SRUKF (Square root unscented Kalman filter) in virtue of square root Kalman filter. The SRUKF calculated the square root of the state covariance instead of state covariance to avoid negative definiteness of the state covariance caused by rounding errors, and ensured the running efficiency and numerical stability of the algorithm to a certain extent. However, to the authors’ knowledge, all the above conventional filtering algorithms can acquire limited accuracy when applied for autonomous navigation on the condition that the spacecraft was stably running on a predefined orbit.

Also, for multiple-sensor data fusion systems, a critical problem is to find an optimal state estimator fusion with the given measurements. There exist two kinds of fusion architectures, as is known, centralized fusion structure and distributed fusion structure, and the former can always provide the optimal fusion results in the least mean-square sense due to the smallest information loss [8], [9], [10], [11]. Therefore, the centralized fusion structure containing two sensors such as earth sensor and star sensor is chosen to realize autonomous navigation.

In this paper, a new information fusion filtering algorithm of ISRUKF is proposed, which integrates both merits of IUKF and SRUKF, and uses the above three measurements. Similar algorithm may have been researched [12], but the proposed ISRUKF method can further estimate measurement noise covariance matrix online to approach the actual system. Meanwhile, considering the systematic computer of spacecraft which needs to carry out many tasks simultaneously, the system has to compromise between estimation accuracy and computational complexity. Finally, a switch-mode filtering algorithm based on ISRUKF and EKF is designed and applied to autonomous navigation system. Semi-physical system validation results show that the switch-mode information fusion filter is suitable for autonomous navigation system which is implemented orbit maneuver with finite thrust. The main contribution of this paper includes two parts (1) the ISRUKF method proposed can fuse three measurement and adapt measurement noise covariance matrix online to approach the actual system and (2) the switch-mode information fusion filter designed can reduce the complexity of the algorithm as well as ensure the accuracy of the estimation.

The rest of the paper is arranged as follows. In Section 2, the system model for autonomous navigation is described; Section 3 proposes the ISRUKF–EKF switch-mode information fusion filtering algorithm. In Section 4, semi-physical simulation is presented to show improvement in accuracy and efficiency of switch-mode information fusion filtering algorithm. Section 5 concludes the paper.

Section snippets

System model for autonomous navigation

Flow of autonomous navigation of spacecraft is showed in Fig.1. By observing stars, the star sensor obtains the star map, which is put on noise elimination and distortion correction processing. Followed by extraction of centroid and calculation of magnitude energy of the star, the star sensor identified the observed stars from the star database and acquires starlight vector information in the spacecraft body coordinate. By observing the earth, the earth sensor obtains the earths disk image,

ISRUKF algorithm

Standard SRUKF has been described in [14], [15]. Here the ISRUKF is put forward based on SRUKF. The ISRUKF has adaptive ability for measurement noise by updating corresponding covariance matrices Rk and raise stability and convergence of the filter by utilizing iterative strategy. The ISRUKF algorithm is given as:

Considering a discrete-time nonlinear dynamic system,xk+1=f(xk,uk,k)+wkyk=h(xk,uk,k)+nkwhere xk is the n-dimensional state, uk is a known control vector, wk represents the process

Simulation and validation results

In this section, the proposed switch-mode filter is applied to the autonomous navigation system, the effectiveness of which can be demonstrated by comparing with ISRUKF and EKF. Then, the results of semi-physical simulation are described as follows.

Conclusion

In this paper, we presented a switch-mode information fusion filtering algorithm based on the ISRUKF and EKF, and used three measurements of navigation sensors simultaneously. This algorithm could adaptively switch between the ISRUKF and the EKF using a switch-mode function and centralized fusion structure. Semi-physical simulation results showed that the switch-mode filter could ensure good robustness and estimation accuracy in case of orbit maneuver with finite thrust, with systematic

Acknowledgements

This work was supported by the National Nature Science Foundation of China (61233004, 61221003, 61074061), the National Basic Research Program of China (973 Program-2013CB035500), and partly sponsored by the International Cooperation Program of Shanghai Science and Technology Commission (12230709600), and the Higher Education Research Fund for the Doctoral Program of China (20120073130006 and 20110073110018). The authors thank the reviewers and editor for their helpful and constructive comments.

References (20)

There are more references available in the full text version of this article.

Cited by (27)

  • Adaptive cubature Kalman filter with the estimation of correlation between multiplicative noise and additive measurement noise

    2022, Chinese Journal of Aeronautics
    Citation Excerpt :

    Considering the fact that traditional Kalman filtering can only deal with linear problems, many nonlinear filtering methods have been proposed. Gaussian filter algorithm is widely used in nonlinear Gaussian systems, which is able to provide a filtering framework for the state estimation, such as Unscented Kalman Filters (UKF)4,5 and Cubature Kalman filters (CKF)6. Comparatively, CKF is recognized to have higher estimation accuracy.

  • A novel method based upon modified composite spectrum and relative entropy for degradation feature extraction of hydraulic pump

    2019, Mechanical Systems and Signal Processing
    Citation Excerpt :

    However, it seems difficult to select appropriate fusion weights for the weighted fusion algorithm [17]. Kalman filtering algorithm lacks of strict filtering functions for nonlinear systems [18]. During the sampling operation in the wavelet analysis fusion algorithm, some sensitive feature information may be lost.

  • The morphological undecimated wavelet decomposition – Discrete cosine transform composite spectrum fusion algorithm and its application on hydraulic pumps

    2016, Measurement: Journal of the International Measurement Confederation
    Citation Excerpt :

    However, it seems difficult to select appropriate fusion weights for the weighted fusion algorithm [16,17]. Kalman filtering algorithm lacks of strict filtering functions for nonlinear systems [18,19]. During the sampling operation in the wavelet analysis fusion algorithm, some sensitive feature information may be lost [20].

  • Performance enhancement of X-ray pulsar navigation using autonomous optical sensor

    2016, Acta Astronautica
    Citation Excerpt :

    There is a tremendous amount of literature on the problem of optical navigation. See [31–34] and the references therein. A sensor capable of performing this task would be similar to the ultraviolet optical sensor, which images the planet and the stars simultaneously in the ultraviolet band [35–37].

View all citing articles on Scopus

VEDANTHAMM 20-JUN-2013 12:21:07 Open Access Yes in EES, but no funding form. OA and CC undone.

View full text