A major issue in the control of dynamical systems is the integration of both technological constraints and some dynamic performance requirements in the design of the control system. We show in this work that it is possible to solve a class of constrained control problems of linear systems by using a reduced-order system obtained by the projection of the trajectories of the original system onto a subspace associated with the undesirable open-loop eigenvalues. The class of regulation schemes considered uses full state feedback to guarantee that any trajectory emanating from a given polyhedral set of admissible initial states remains in that set. This set of admissible states is said to be positively invariant with respect to the closed-loop system. We also address the important issues of numerical stability and complexity of the computations.