A new design methodology that yields robust controllers for globally or locally linearized model is introduced. The linear models are used together with eigen-structure assignment to select suitable dynamics for the system. The selected eigen-structure is assigned to a linear regulator with prescribed relative stability. The satisfaction of state inequality constraints is checked using a published method due to Kiendl. After the designs are completed, their robustness is assessed using a new measure of robustness. The process is repeated until satisfactory robustness, relative stability, etc. are achieved. The design methodology is illustrated using a 2-DOF robotic manipulator example. The example shows that the final design selected must be a compromise between robustness, relative stability, gain magnitude and other considerations.