Mark Pittelkau, Solers, Inc.
Keywords: Space Situational Awareness, Multiple Sequential Hypothesis Testing, Data association
Abstract:
A space object from a surveillance catalog of space objects that is predicted to be in the field of view of a tracking sensor may not be detected by a tracking sensor because of viewing conditions, or because either the estimated orbit of the space object has large error or the catalog object does not actually exist. In these two latter cases we call such a catalog object an invalid object (a false or lost track). Identification of invalid catalog objects is an essential function for maintenance of a space surveillance catalog. An invalid catalog object is not likely to be associable with any measurements (observations) in a sequence of data collects from tracking sensors. The sequential probability of validity over multiple frames of data is cumulative evidence of whether the catalog object is valid or invalid. The catalog object is deemed to be invalid when the sequential probability of validity is sufficiently close to zero. The single-frame and sequential probabilities of validity of a catalog object are determined by first computing the maximum likelihood association of the tracks (the estimated orbits) and observations, and then by updating a sequential likelihood ratio test for each track. The probability that each track is valid is computed from likelihood ratios. The algorithm is simpler and computationally faster than the general Multiple Hypothesis Testing (MHT) algorithm that associates multiple measurements and multiple objects over multiple frames of data. The sequential data collects do not have to correspond to the same set of catalog objects, and the data collects can be separated in time and can come from tracking sensors that are geographically separated, have different views, and that report different types of measurements (angles, range, range rate, position vectors, or position and velocity vectors). The measurement data can include feature (attribute) data. The only condition is that the catalog object under consideration is predicted to be in or near the field of view of the tracking sensor in multiple data collects; whether or not there are any associable measurements in the data collects is determined by the algorithm.
Date of Conference: September 20-23, 2016
Track: SSA Algorithms