An Adaptive Fuzzy State Noise Driven Extended Kalman Filter for Real Time Orbit Determination
- Paper number
IAC-07-E2.1.03
- Author
Mr. Rohit Garhwal, Indian Institute of Technology, India
- Coauthor
Mr. Abhishek Halder, Indian Institute of Technology, India
- Coauthor
Prof. Manoranjan Sinha, Indian Institute of Technology, India
- Year
2007
- Abstract
The problem of orbit determination is of considerable significance for the early initiation of action for on-board operations including satellite tracking and control. Generally Extended Kalman Filter (EKF), which is a suboptimal nonlinear implementation of linear Kalman filter, is employed for the real time orbit determination. However, the divergence of the EKF can not be ruled out or at least a poor convergence may creep in even after employing various methods to make it adaptive by injecting noise. The divergence may occur due to errors in modeling the system, finite precision arithmetic and associated truncation/round-off errors and large errors can be attributed to a priori estimate and covariance. The artificial noise injection used for making the state covariance matrix positive definite, may not lead to proper convergence due to the problems mentioned above. Therefore, any effort to model this filter based on just artificial noise injection may not make the filter robust. In this paper a fuzzy state noise driven adaptive EKF has been proposed for orbit determination. Here fuzzified state residuals are used for constructing the noise model rather than observation residuals. A fuzzy damping analogy has been applied to construct the state noise covariance matrix. Each of the diagonal elements is constructed using the fuzzified corresponding state vector component residuals and their derivatives. The formulation is elegant enough to drive the EKF towards a faster convergence with higher fidelity, yet simple enough to make it suitable for real time implementation. A comprehensive simulation on PSLV-C1 data has been carried out using various existing methods and the method proposed in this paper. It is shown that the fuzzy state noise driven EKF performs much superior to other models, making it a better choice for the real time application with a greater confidence.
- Abstract document
- Manuscript document
IAC-07-E2.1.03.pdf (🔒 authorized access only).
To get the manuscript, please contact IAF Secretariat.