A NEW SWITCHING CRITERION FOR THE DUAL-MODE ORBIT PREDICTOR

The estimation of the orbital states of a satellite is com-plicated by the fact that a realistic model of satellite dynamics is nonlinear and so is the observation model. Among the various schemes for nonlinear state estimation, it was found in previous work that a combination of invariant imbedd-ing with stochastic approximation provides accurate estimates. The method consists of starting estimation with invariant imbedding and switching over to stochastic approximation at a
suitable linstant of time. A eritetion that governs the switching instant was devised. It was developed by using the statistical properties of the innovations sequence. The "degree of whiteness" of the in-novations sequence was used as an indicator of the overall performance of the estimator. However, when the number of observations is small, this switching criterion fails to give acceptable results. In this paper, a new switching criterion for the dual - mode orbit predictor is presented. This criterion is based on a modification to the measure of whiteness of the innovations sequence. Results of simulation using data supplied by the Communications Research Centre of Canada are included and demonstrate the effectiveness of the proposed criterion when moderate sample sizes are available.