Marek Moeckel, SERC, Ltd., James Bennett, SERC, Ltd., Enrico Stoll, Technische Universitat Braunschweig, Kefei Zhang, RMIT University
Keywords: Analytical Propagation, GPU-Computing, High Performance Computing
Abstract:
Orbital propagation is a key element in many fields of space research. Over the decades, scientists have developed numerous orbit propagation algorithms, often tailored to specific use cases that vary in available input data, desired output as well as demands of execution speed and accuracy. Conjunction assessments, for example, require highly accurate propagations of a relatively small number of objects while statistical analyses of the (untracked) space debris population need a propagator that can process large numbers of objects in a short time with only medium accuracy. Especially in the latter case, a significant increase of computation speed can be achieved by using graphics processors, devices that are designed to process hundreds or thousands of calculations in parallel. In this paper, an analytical propagator is introduced that uses graphics processing to reduce the run time for propagating a large space debris population from several hours to minutes with only a minor loss of accuracy. A second performance analysis is conducted on a parallelised version of the popular SGP4 algorithm. It is discussed how these modifications can be applied to more accurate numerical propagators. Both programs are implemented using a generic, plugin-based software architecture designed for straightforward integration of propagators into other software tools. It is shown how this architecture can be used to easily integrate, compare and combine different orbital propagators, both CPU and GPU-based.
Date of Conference: September 20-23, 2016
Track: SSA Algorithms