The purpose of this section is to define a cost or penalty associated with an orbit identification. This penalty takes into account the uncertainty of both orbit determinations, but the uncertainty is not easy to define as it depends upon the timing, quantity, and quality of the astrometric observations. There are, however, rigorous and quantitative ways to describe the uncertainty of orbits based upon the normal/covariance matrix formalism, provided the orbits have been computed as a solution of a least squares fit. In the next subsection we recall the notation and terminology used to describe a least squares solution. In the following subsections we describe the linear identification algorithm in the unrestricted and in the restricted form, respectively.