In this paper, linear over-actuated systems are considered, i.e., systems with more control inputs than regulated outputs. For these systems, an entire family of input functions and/or state trajectories is compatible with a prescribed reference for the output. The possibility of dynamically selecting on-line the most suitable of these trajectories according to certain performance criteria, leads to the so-called dynamic input allocation problem. In this paper we present, within the general framework of the robust output regulation problem, the design of dynamic input allocators for linear uncertain system, on the basis of two selected performance criteria of practical interest. Specifically, we aim at on-line minimization of the weighted energy and the amplitude of the control input, criteria that are especially suitable for input energy optimization and to deal with saturation limits. Stability and robustness of the ensuing closed-loop system are proved. An example is provided in support of the methodological findings.