Information Theoretic Criteria for Observation-to-Observation Association.

摘要

There are three types of data association problems. The first is the observation-to-track association (OTTA) problem, where given an observation with some known measurement statistics and a set of existing candidate (uncertain) resident space object (RSO) tracks the analyst seeks to associate each observation with a unique track (or none). The second association problem is where we have multiple tracks at different time instances and wish to determine whether any of the tracks belong to the same RSO. This is the track-to-track association (TTTA) problem. The final association problem is where we are given a set of observations at different time instances and wish to determine which of these observations were generated by the same RSO. This is the observation-to-observation association (OTOA) problem. The focus of our paper is the OTOA problem. In this paper, we tackle the OTOA problem by using an appropriate initial orbit determination (IOD) method as well as criteria from information theory. The two main criteria we use in this paper are mutual information and information divergence. We demonstrate how these two criteria can be used within an unscented transform framework as well as with a particle-based approach. The information theoretic solutions described in this paper can be adjusted to address the other (OTTA and TTTA) association problems, which will be the focus of future research. We will demonstrate the main result in simulation.