Performance survey of robust pole placement methods

The classic problem of robust pole placement for linear time invariant systems via state feedback has been studied for several decades, and involves obtaining a gain matrix that will assign a certain desired set of closed-loop poles, while also providing a robust eigenstructure that is insensitive to uncertainties in the system matrices. There are several ways of measuring the robustness of the eigenstructure, and numerous methodologies have appeared in the literature to address the problem. In this paper, results from extensive experiments comparing the performance of a number of methods-including variations on two methods previously proposed by the present authors-against a variety of robustness measures are reported. The size of the matrix gain, runtime and accuracy of the pole placement achieved by each method are also compared. The results show some notable differences between the methods surveyed.