Particle filters (PFs) offer the possibility of addressing many unsolved problems in orbit determination and prediction. This work builds on an existing GSRP (Graduate Student Researchers Project) effort to incorporate a particle filter into GSFC’s (Goddard Space Flight Center’s) Orbit Determination Toolbox (ODTBX). Extensions were investigated that partition the particles into subsets based on a priori errors, measurement noise, process noise, and maneuver execution errors. Parallel computing was used to efficiently implement a sufficiently large number of particles to begin solving significantly non-Gaussian estimation problems.
The work builds on ODBTX’s inherent capability to distribute computations over multiple computational cores, which will prove complementary to the structure of the PF, since computations involving each particle can be assigned to their own core.
The PF estimator added to ODTBX, when running on a multicore server, will enable GSFC navigation analysts to tackle previously daunting navigation challenges, such as simultaneously determining the orbit around, and characterizing the properties of primitive solar system bodies. It will provide a benchmark for the current suite of orbit estimators, uncovering any shortcomings that arise from Gaussian assumptions in the current estimators.
This work was done by John A. Gaebler, Alinda Mashiku, and Russell Carpenter of Goddard Space Flight Center. GSC-16608-1