Orbit Determination and Orbit Uncertainty Estimation using Neural Network Based Modified State Observer

In order to improve the accuracy of the dynamical model used in the satellite orbit determination, a novel technique called modified state observer, is presented for the estimation of uncertainties in nonlinear dynamics and, in addition, providing the estimate of the system states. The observer structure contains an online neural network whose outputs are the uncertainties in the system. To demonstrate the utility of the presented algorithm for this class of problem, the observer is applied in two cases: estimating the uncertainty caused by just the J2 perturbation and estimating all dynamical perturbations and satellite state of a GPS satellite. Simulations are presented, and results show that the observer can accurately estimate the perturbation and the satellite states.