Elimination of latent variables in behavioral systems: a geometric approach

This paper considers computational aspects of the problem of elimination of variables. More precisely, the problem of elimination of latent variables in models is addressed in the behavioral framework. In earlier contributions on this theme, the model classes of infinitely smooth and square integrable linear time-invariant systems have been considered. For both system classes, conditions for elimination of latent variables have been derived. However, efficient computational scheme to eliminate distinguished variables from equations have received less attention in the literature. This paper addresses sufficient conditions for the elimination problem where algebraic state space operations are used. We provide a constructive and computable algorithm for the elimination of latent variables.