Estimation of the Local Attitude of Orbiting -Spacecraft

Abstract

Precise knowledge of spacecraft attitude is of interest for satellites nominally pointed at the earth, when aiming telescopes, cameras, and antennas, and as an essential ingredient in an active attitude control system. In earth orbit an horizon sensor can measure pitch and roll angles but cannot measure yaw angle since yaw motion does not change the vehicle's attitude with respect to the horizon. A gyro can sense only changes in yaw attitude so that information on initial yaw is required; furthermore, the measurement accuracy decreases with time (gyro drift). By combining measurements from an horizon sensor and two rate-integrating gyros (in roll and yaw), yaw angle can be estimated without loss of accuracy as time increases. Such a device is called an orbital gyroeonpass; its logical structure follows directly from Knlman-Bucy filter theory if the measurement uncertainties are approximated as white noise, and only the kinematic equations of motion are used.The measurement uncertainty of rate-integrating gyros is discussed and it is suggested that, for long periods of operation, a constant bias plus exponentially-correlated noise is a better approximation than white noise or a random walk process. The logical structure of the optimal filter is deduced for gyros with additive exponentially-correlated noise, using only the kinematic equations of motion. Using current estimates for measurement uncertainty, it is shown that this filter gives estimates of roll and yaw angle with steady-state RMS errors that are lower than the orbital gyrocompass filter by a factor of two to three. The improvement comes largely from the estimation of yaw-gyro drift. Although the improved filter is designed assuming exponentially-correlatod noise, it also estimator, a constant bias drift reasonably well'.In the derivation of the filters, rate gyro measurements are assumed. It is shown that the filters can be implemented using rate-integrating gyros which are substantially more accurate than rate gyros. The steady state filter appears to be adequate for many missions so that time-varying gains are not needed.A more complicated estimator is discussed which uses both the kinematic and dynamic equations of motion. Surprisingly, this estimator often does not give improved estimates of roll and yaw angle.