Applications of square-root information filtering and smoothing in spacecraft orbit determination

Abstract

The JPL (Jet Propulsion Laboratory) orbit determination software system is a set of computer programs developed for the primary purpose of determining the flight path of deep-space mission spacecraft in NASA's planetary program and highly elliptical orbiting spacecraft in Earth orbit. The filtering processes available within the JPL orbit determination software are discussed, and several examples are presented. In particular, solutions obtained by the square root information filter (SRIF) using Bierman's estimation subroutine library (ESL) are discussed and compared with the solutions obtained by the singular value decomposition (SVD) technique. It is concluded that the SRIF filtering and smoothing algorithms are efficient and numerically stable for well-conditioned systems. The use of Bierman's ESL simplifies the task of maintaining the orbit determination software by providing efficient, tested filtering tools. For solving a large well-conditioned system (rank >120), SRIF is approximately four times faster than SVD; however, for solving an ill-conditioned system, SVD is recommended. >