Description
One of the challenges of orbital prediction calculations is the propagation of uncertainty. A numerical modelling technique is presented that combines the low computational overhead and continuity of a low-fidelity Keplerian Monte Carlo simulation with the accuracy of a high-fidelity orbital dynamics simulation at key points.
Effective decision making in space domain awareness (SDA) relies on accurate estimates of uncertainty, but the non-linear nature of orbital mechanics and numerous small perturbations on objects in planetary orbits makes uncertainty propagation difficult. A modelling technique has been developed that approaches this problem with the goals of accuracy, computational efficiency, and algorithmic simplicity. By using a multifidelity approach, computationally intensive operations are limited to updates of a Monte Carlo ensemble at key points.
The model starts with an ensemble of points to be propagated using low-fidelity Keplerian equations in a Monte Carlo simulation. The ensemble is approximated by a Gaussian mixture model that is defined by points along each eigenvector of each Gaussian component. These points are propagated in time using both the computationally cheap low-fidelity model and a resource-intensive high-fidelity model that simulates detailed orbital perturbations. By comparing the low and high-fidelity Gaussian mixture points after propagation, the low-fidelity Monte Carlo ensemble is updated and used as the basis for an new Gaussian mixture model before the process is repeated.
Testing of this technique demonstrated excellent agreement with a full high-fidelity Monte Carlo simulation with a time savings of two orders of magnitude for objects in low Earth orbit.
| I am the presenting author | Yes |
|---|