Statistical approach to robust-performance control synthesis for systems with uncertain physical parameters

This paper introduces a new design approach integrating robust control and quality control to synthesize robust-performance controllers for systems with uncertain physical parameters. We exploit classic statistics and related inference theories to experimental design that uses orthogonal arrays to directly achieve time domain specifications, which is difficult for approaches based on frequency-domain synthesis such as norm-bounded optimization schemes. In addition, by exploiting analysis of variance (ANOVA), the interrelation between specifications and each design parameter can be studied quantitatively. In comparison with existing optimization schemes, our proposed scheme requires less numerical computations without the trouble of initial guess, and without differentiation operation. Systematic procedures for this quality-oriented control design are outlined in this paper.