GYRO-STELLAR ATTITUDE ESTIMATION CONSIDERING MEASUREMENT NOISE CORRELATION AND TIME-VARIANT RELATIVE SENSOR MISALIGNMENT
- Paper number
IAC-08.C1.7.4
- Author
Dr. Stefan Winkler, EADS Astrium GmbH, Germany
- Coauthor
Dr. Wilhelm Gockel, Germany
- Coauthor
Dr. Georg Wiedermann, Germany
- Year
2008
- Abstract
- 1 Overview
This paper deals with satellite attitude determination based on optimal fusion of measurements from an Inertial Measurement Unit (IMU) and star trackers (STR) by stochastic filters (gyro-stellar estimation). Against commonly made simplifications like white measurement noise and time-constant sensor alignment, the paper focuses on attitude accuracy improvement taking star tracker measurement noise correlations and relative IMU/STR distortion into account. For this, different approaches are presented and evaluated from a real satellite project point of view. The results presented are part of the current phase-B2 Attitude and Orbit Control System development activities for the European Sentinel-2 satellite.
2 MotivationCurrent commercially available high-precision STRs provide attitude accuracy in the order of 1 arcsec (8 arcsec) (1σ) in cross-boresight (boresight). This performance can be improved by a least-squares solution using more than one STR with optimal boresight alignment. However, the attitude performance is still limited by alignment restrictions (e.g. due to blinding) and time-correlated noise components. To overcome these effects, a gyro-stellar estimator can be used.
Often, STR measurement noise is treated as white noise within the filter design. This rises the question:
- What is the attitude performance improvement vs. the additional computational load when noise correlation is considered within the filter?
STRs and IMU are usually mounted close together to avoid time-variant relative misalignment. Often, like for Sentinel-2, it is preferred to distribute the sensors within the satellite leading to time-variant STR/IMU misalignment due to thermal distortion. Regarding high-accuracy attitude determination requirements (e.g. Sentinel-2: <10 µrad, 2σ), the following questions rise:
- What attitude error is induced by unknown time-variant STR/IMU misalignment?
- What is the achievable attitude performance considering time-variant misalignment within the filter?
These questions will be answered in this paper.
3 Attitude Estimation Algorithm and ResultsThe developed gyro-stellar estimation algorithm is based on an extended Kalman filter.
Several methods to consider time-correlated measurement noise within a Kalman filter exist (e.g. state augmentation, measurement differencing). In order to consider the time-correlated noise components of a STR measurement, state augmentation was used because it provides reliable results without time-consuming filter tuning. Details on quantitative attitude improvement will be given and discussed in the paper.
Considering 100 µrad time-variant misalignment amplitude per axis increases the attitude error by 82% per axis despite filter covariance tuning. Considering time-variant misalignment by an appropriate dynamical model within the filter increases the attitude error by only 6% per axis. This result clearly reflects the capability of the approach to be presented. A detailed presentation and discussion of the proposed approach will be given in the paper.
- Abstract document
- Manuscript document
IAC-08.C1.7.4.pdf (🔒 authorized access only).
To get the manuscript, please contact IAF Secretariat.