Extended Kalman filtering and nonlinear predictive filtering for spacecraft attitude determination

Demands in the design and manufacture of spacecraft are changing. Smaller, lower-cost spacecraft are becoming increasingly more desirable, while the required spacecraft performance is increasing (consider the space telescope, which is required to accurately track a star). Use of cheaper hardware would lower the spacecraft performance, but this effect may be offset by the use of more sophisticated controller and estimation schemes. This paper compares the Extended Kalman Filter (which has traditionally been used for spacecraft attitude determination) with a relatively new filter, the Nonlinear Predictive Filter. It is shown that tinder certain circumstances the Nonlinear Predictive Filter outperforms the Extended Kalman Filter, but Mat the requirements for these circumstances (namely the use of rate sensors) conflict with the robustness requirement with respect to initial estimate error.