Background

On large spacecraft, high precision attitude determination can be performed using high performance sensors such as star trackers and inertial grade gyros. On nano satellites, however, such instrumentation is not available due to small budgets for mass, volume and power. Furthermore, any attitude sensor that is used will take away valuable budget from the science payload. Consequently it is desirable to be able to perform attitude determination on a nano satellite using few, small sensors.

A candidate minimal sensor set would be the spacecraft’s own solar panels. As all satellites require electrical power, they already have solar panels. Thus, no additional components are being added to the spacecraft, and no additional mass, volume or power budget is being used. From differential solar panel currents, an estimate of the solar vector can be obtained, and this can be used to perform attitude determination.

O/OREOS - NASA Ames Research Center RAX-1 - University of Michigan

Attitude determination can still be performed on a nano satellite using only a measurement of the solar vector, but an estimator that incorporates an accurate spacecraft attitude dynamics model is required. The same low mass and power budgets that negate the use of high performance attitude sensors also encourage the use of passive magnetic stabilization systems in nano satellites.

Problem

When the magnetic components are installed in a spacecraft the magnetic properties will change due to interactions both with similar material installed in close proximity, and also to other spacecraft components. Prelaunch testing in the laboratory of a complete system is hampered by the very small torques produced, and the difficulty in maintaining a magnetically clean environment. The differences between laboratory measured properties and the effective installed properties have been substantial enough to cause problems with attitude determination, with previous nano satellite missions observing disagreements between pre-launch simulations and observed attitude profiles. In many cases, the observed bus measurements cannot be reproduced using the pre-launch dynamics model, as demonstrated in the figure below, which compares bus solar data to that derived from simulation.

Mismatch between observed data and simulated data - attributed to poor dynamics model

The errors in the assumed dynamics model mean that attitude estimation cannot be performed, either using batch or recursive methods.

Batch Attitude Estimation Error Recursive Attitude Estimation Error

Approach

Before attitude estimation can be performed using only an estimate of the solar vector, the attitude dynamics must be calibrated. This will be done using actual flight data. As attitude dynamics are non linear and non convex, estimation will be performed using a batch method that has at its core a non-convex optimization. 

Associated Personnel

Last modified Fri, 22 Nov, 2013 at 10:20