Fig. 9-16. RAE-2 Data Selection Based on an A Priori Attitude inspection of Fig. 9-16(a) suggests the nature of the anomaly. Nearly all of the AOSs are spurious and only the lunar presence at 25 deg < y<35 deg resulted in valid LOSs at the lunar terminator. For RAE-2, these displays, which utilized an a priori attitude and interactive graphics, were essential for attitude determination and maneuver planning during the early portion of the mission.

The attitude determination process is more complex than described above whenever data selection based on an a priori attitude is not sufficiently accurate. This occurs, for example, in the presence of smoothly varying systematic anomalies in which some of the data are clearly invalid but presumably "valid" and "invalid" data run smoothly together. Attitude determination in the presence of such errors requires iterative processing to obtain successive attitude estimates. The general procedure for this is as follows:

1. Discard "obviously" bad data (in addition to the rejection of random errors as described in Sections 9.2 and 9.3).

2. Use the remaining data to estimate the attitude as accurately as possible.

3. Use the new attitude estimate to reject additional data (or recover previously rejected data) as appropriate.

4. Iterate until a self-consistent solution has been obtained, i.e., when step 3 makes no change in the set of selected data.

This procedure does not establish that the final attitude estimate is correct, or that the data selection has been correct. It is also possible that the iterative procedure will not converge—it may eventually reject all the data or oscillate between two distinct data sets. This method can at best obtain an attitude solution which is consistent with the data selection process. Therefore, whenever problems of this type are encountered, it is important to attempt to find the physical cause or a mathematical model of the data anomaly to provide an independent test of whether the data selection is correct.

The central problem of the above iteration procedure is the data rejection in step 3. Operator judgment is the main criterion used, both because general mathematical tests are unavailable and because the anomaly is usually unanticipated. (Otherwise it would have been incorporated as part of the attitude determination model.) Tables of data are of little or no use for operator identification of systematic anomalies; therefore, data plots are normally required. Four types of data plots are commonly used for this purpose:

1. Plots of raw data

2. Plots of deterministic attitude solutions obtained from individual pairs of points within the data

3. Plots of residuals between the observed data and predictions from a least-squares or similar processing method based on the entire collection of data

4. Plots comparing directly the observed data and predicted data based on the most recent attitude estimate

In practice, the author has found the fourth type of plot to be the most useful in defining the boundary between valid and invalid data. To illustrate the use of various plot types and the process by which anomalous data are identified, we describe the data selection process which was used to eliminate the "pagoda effect" identified in SMS-2 data.

The SMS-2 Pagoda Effect. The Synchronous Meteorological Satellite-2, launched from the Eastern Test Range on Feb. 6, 1975 (Fig. 1-1), was the second test satellite for the Geostationary Operational Environmental Satellite series used by the U.S. National Oceanic and Atmospheric Administration to provide daily meteorological photographs of the western hemisphere and other data. During the transfer orbit to synchronous altitude, attitude data were supplied by two Sun sensors and five body-mounted, infrared horizon sensors. As illustrated in Fig. 1-6, each horizon sensor sweeps out a conical field of view or scan cone. Because the spin axis was nearly fixed in inertial space, the scan cone of a single Earth sensor encounters the Earth during one or two segments of the spacecraft orbit and moves across the disk of the Earth as the spacecraft moves. As shown in Fig. 9-17(a), a major anomaly, called the pagoda effect, occurred in the Earth data [Chen and Wertz, 1975]; this is most easily seen in die sharp upturn of Earth-out data as the

0 0

Post a comment