Abstract This work presents a novel scheme for robust constrained control of multivariate over-actuated systems. It combines two well-known control techniques for systems with limitations: control allocation and reference governor. With allocation of control, the number of manipulated variables is reduced to only three aerodynamic moments around the aircraft body axes, simplifying considerably the subsequent control synthesis. To improve the robustness of the operation that converts virtual into real control inputs, the robust set of attainable moments is defined, which accounts for several uncertainties in the vehicle control effectiveness. This set is then treated as the main system constraint in a new form of reference governor. The virtual input is calculated based on feedback of estimated states and feedforward of the steady-state aerodynamic moment, but the reference is governed in order to guarantee constraint satisfaction. A sequence of off-line polyhedral projections eliminates the need of on-line optimization for the computation of control actions. The complete control method is then illustrated with two design cases. Simulations are conducted with two fighter aircraft models, in the presence of uncertainties on control and stability derivatives, as well as large variations in flight conditions. Results demonstrate satisfactory performance while tracking demands of large amplitude, with proper and admissible distribution of control effort.