SIM will need to build its own precision reference frame because no other mission has ever before measured star positions with such accuracy. The solution is to create a network of relative positions over the entire sky. By tying together sets of relative measurements of star separations, we can create a rigid reference frame. The system is much like laying out a lattice of stiff yardsticks into a frame.
The astrometric grid is a model of the positions and motions of a particular collection of astronomical objects. These objects will mostly be stars, but some will be quasi-stellar objects (or quasars) that provide an extragalactic anchor for the astrometric grid. These grid objects will be chosen to uniformly cover the entire sky.
SIM will observe the grid a few times a year; with the data it collects, scientists will be able to calibrate the instrument and compute the positions and motions of the objects in the grid. This calibration is performed by
fitting the grid model to the set of grid observations that SIM has collected. This calibration can then be used to perform science with SIM.
At SIM's level of accuracy, there are no "fixed" stars. All stars will move in a well-resolved fashion in parallactic ellipses, due to the motion of Earth around the Sun and the true movement of the stars through space, proper motion.
For some observations a local grid will be sufficient. The detection of planetary companions does not require the full knowledge of absolute positions but can rely on changes in position relative to reference stars close to the object of interest.
Once sufficient grid observations have been made to build up a reliable grid model, observations of grid objects made in conjunction with observations of science objects allow SIM to perform accurate astrometry on science targets. SIM observes grid objects and uses the grid model to solve for interferometer parameters (baseline length and orientation). This process provides the proper calibration for observations of science objects.
It is important to note that it is not necessary for the grid solution to be completed before science operations begin -- instead combined grid and science measurement scenarios are made, then properly interpreted in the
context of the grid solution later.
SIM Science Investigations