Disturbance localization by measurement feedback for linear periodic discrete-time systems (Q1108245)

This paper deals with systems described by equations of the following type: \[ x(k+1)=A(k)x(k)+B(k)u(k)+M(k)d(k),\quad y(k)=C(k)x(k),\quad z(k)=D(k)x(k) \] where x is the state, u is the control input, d is the unmeasurable and unknown disturbance input, y is the measured output, z is the regulated output and the matrices A(k), B(k), C(k), D(k) and M(k) are periodic with period \(\omega\). Necessary and sufficient conditions are given for the existence of an \(\omega\)-periodic dynamic compensator, having y as input and u as output, such that the output z of the over-all control system is disturbance decoupled (i.e. not influenced by the disturbance d), in three different cases: without and with an additional requirement of output or state dead-beat control in the free response (called Problems DLPM, DLPOCM and DLPSCM, respectively). Algorithms for the effective design of the compensator, if it exists, are also provided. Both the conditions and the design procedures are based on the use of some geometric concepts, here introduced (or recalled) for this class of systems, studied, and characterized for control purposes through proper algorithms; namely, the \(\omega\)-periodic notions of (A,B)-invariant (or controlled invariant) subspace, (C,A)-invariant (or conditionally invariant) subspace, inner controllable subspace (i.e., an (A,B)- invariant subspace whose states are controllable to the origin with a trajectory contained in the subspace), outer controllable subspace (i.e. an (A,B)-invariant subspace into which the state of the system can be steered for any initial state), outer reconstructible subspace (i.e., a (C,A)-invariant subspace consisting of the unknown-input unreconstructible states corresponding to an input entering through some \(\omega\)-periodic matrix E(k)), and (C,A,B)-pair of subspaces (i.e., a pair of a (A,B)-invariant and (C,A)-invariant subspaces, the former containing the latter). The conditions are stated in terms of the above subspaces as follows: (i) Problem DLPM is solvable iff \(V*(k)\subset V^*(k)\) for all integers k; (ii) Problem DLPOCM is solvable iff \(V^{ou}*(k)\subset V^*(k)\) for all integers k, and \(V^*(k)+X_ c(k)=X\) for an arbitrary integer k; (iii) Problem DLPSCM is solvable iff \(V^{ou}*(k)\subset V^*_{in}(k)\) for all integers k, and the system is controllable and reconstructible, where X is the state space, \(X_ c(k)\) is the subspace of the controllable (to the origin) states at time k, \(V^*(k)\) (V*(k)) and \(V_{in}^*(k)\) \((V^{ou}*(k))\) are the largest (the least) \(\omega\)- periodic (A,B)-invariant ((C,A)-invariant) subspace, and, respectively, inner controllable (outer reconstructible) subspace contained in KerD(k) (containing ImM(k-1)). Algorithms are also given for the effective computation of such subspaces; through them these conditions can be easily checked. The design procedures are pointed out, and are clarified by means of a numerical example.