Rejection of anomalous observations when determining spacecraft trajectories

The majority of techniques used to reject coarse anomalous observation errors when tracking space vehicles is based on procedures which smooth the observations with a polynomial of some degree. A more economical method which precludes the necessity of preliminary smoothing is possible and eliminates the problem of the optimal choice of the degree of the smoothing polynomial. The major task of processing an observational sample consists in discriminating the set of measurements which does not contain anomalous data. This problem is solved here in three steps, involving the generation of an initial sample of measurement residuals, followed by the discrimination from the remaining residuals of new measurement residuals which can be included in the initial set and then applying particular criteria to split the initial sample into one set of valid measurement residuals and a set of anomalous residuals. Three specific cases considered are: (1) the initial measurement error residuals have uniform distributions; (2) the errors have a normal distribution; and (3) the measurement errors have a normal distribution only in a narrow range.