One requirement for structural systems is that a local damage must not cause disproportionate conse- quence. This property is termed robustness and demands the presence of alternative load paths. The existence of alternative paths can be assessed with the redundancy matrix that assigns a redundancy value to every structural element. In this work robustness is evaluated as a purely structural attribute and can be evaluated independently of external loads. The focus is on the additional displacements after local damage occurs. Employing the Sherman-Morrison-Woodbury formula from reanalysis methods on the difference between the stiffness matrix of the damaged system and the stiffness matrix of the undamaged system leads to the so-called amplification matrix. This matrix directly relates the nodal displacements of the undamaged structure to the additional displacements that occur after local damage. It incorporates the previously de- fined redundancy values. A singular value decomposition of this matrix yields the maximum possible amplification of the displacements after a given damage. The distribution of the additional displacements for each element’s damage, i.e. the shape of the vector of additional displacements, can also be derived. These derivations enable comparison of various damage scenarios and identification of critical load cases for each scenario. The objective of the robustness optimization is to minimize the maximum amplification factors that were just introduced. Optimization can considerably increase the robustness of a structural design. In the talk the computational framework for assessing and subsequently optimizing structural robustness will be illustrated, with a focus on the question how robustness can be effectively increased without using any information about prospective load cases.