Optimal model reduction of discrete-time descriptor systems

From MaRDI portal





The SIMO decriptor system \(G(z)=C(zE-A)^{-1}b\) has a unique expansion in \(0<|z|\leq\delta\): \(G(z)=\sum_{k=-\mu}^\infty C\phi_k b z^{-(k+1)}\). Such a system is equivalent to two disconnected systems: a causal one \((I,J,b_1,C_1)\), and a non-causal one \((N,I,b_2,C_2)\), where \(N\) is nilpotent with index \(\mu\geq 0\). A reduced system is obtained by minimizing \(E=\sum_{k=-\mu}^\infty\text{ Tr}(\Delta_k\Delta^T_k)\) with \(\Delta_k=C\phi_kb-C_r\phi_k^rb_r\), where the index \(r\) refers to the reduced system. The proposed method consists in finding a parametrization of the causal subsystem using a Routh array technique in a canonical representation. Then, \(E\) is minimized with parameters \(\alpha\), the vector of denominator coefficients in the causal system and the elements in \(C_r\). A gradient based method is used. Explicit expressions for \(E\) and the gradient are derived. Although the analysis is made for SIMO systems, the technique can immediately be generalized to MIMO systems.











This page was built for publication: Optimal model reduction of discrete-time descriptor systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3153799)