State interpolation for on-board navigation systems

Abstract

Common concepts for autonomous on-board navigation systems rely on the numerical integration of a spacecraft trajectory between subsequent measurements of a navigation sensor such as GPS. In combination with a Kalman filter, a predicted state vector becomes available at discrete, but not necessarily equidistant time steps. When used for real-time attitude control or geo-coding of image data, the on-board navigation system has to provide continuous dense output at equidistant time steps, which usually conflicts with the natural stepsize of the relevant integration methods and the non-equidistant measurement times. To cope with this problem, the integrator has to be supplemented by an interpolation scheme of compatible order and accuracy. After presenting a representative formulation of an on-board navigation system and deriving related timing and accuracy requirements, suitable Runge–Kutta methods and associated interpolants are selected and evaluated. Promising results are obtained for the classical RK4 method in combination with Richardson extrapolation and 5th-order Hermite interpolation. The 5th-order Fehlberg method with interpolation due to Enright and, for drag-free scenarios, the 5th-order Runge–Kutta–Nystrom method with 5th-order Hermite interpolation provide a good performance in terms of position interpolation. However, as both methods exhibit significant errors for the velocity interpolation, they are not recommended for use with the outlined navigation filter.