抄録
This paper proposes a method for designing control systems based on the decomposition of time scale, and discusses its property by the singular perturbation theory. The system treated in the paper is a multivariable linear time-invariant system consisting of subsystems S1 and S2 such that S1: x=A11x+A12z+B1u1, S2: z=A21x+A22z+B2u2.The following procedure is proposed to regulate the state x. i) Find the controllers' gains K1, K2, so that S1 with z=K1x, u1=K2x is stable and satisfies the pre-assiged specification, i.e., it satisfies x*=(A11+A12K1+B1K2)x*=Ã1x*, with Ã1 stable, and also find K such that Re λi [B2K]<0. ii) And then construct the feedback system by applying u1=K2x and u2=(1/ε)K(z-K1x), where ε>0. Such a design method is achieved if (A11, [A12 B1]) is stabilizable and rank B2=dim z. It is shown that the feedback system is stable and the trajectory x tends to x* as ε→0. As the second part of the paper, the case where the gains K1, K2, and K are determined by the solutions of LQR problems is considered. It is shown that such gains give a suboptimal solution to the LQR problem of the total system with an appropriate cost function. Based on the LQR theory, the robust stability is discussed. Finally, a numerical example is given for illustration.