A multipropagator approach to real-time orbit simulation

Error control in numerical integration schemes typically focuses on selecting integration stepsizes according to how rapidly the integrand is changing over some interval of its domain. The validity of this approach rests on the assumption that the output error may be reduced to an arbitrarily small value by selecting an arbitrarily small time interval of integration. In the context of real-time simulation of spacecraft orbits, however, performance constraints may impose an effective lower limit on orbit integration stepsizes. Clearly, overall performance requirements dictate that consideration be given to selecting not only an appropriate integration interval but also a propagation algorithm that is itself appropriate to the nature and magnitude of the forces imposed on the simulated vehicle. Here the behavior of the output error from two typical numerical integration algorithms is examined for several test orbits of various types, and the issue of defining reliable, objective, quantitative criteria for orbit propagator selection is addressed. An example of the catastrophic failure of a standard finite difference integration algorithm in the presence of strong atmospheric drag is presented. Rehabilitation of the failure by automatic propagator selection is exhibited for the case of a highly eccentric orbit decaying under intermittent atmospheric drag.