The separation loop as an architectural distinction between classical and modern control systems

It is argued that straightforward generalization of the architecture of classical single-input single-output (SISO) control systems to multiple-input multiple-output (MIMO) systems introduces several issues which are avoidable in a more appropriate architecture. In fact, active implementation of the architecture implicit in Kalman's concepts of controllability, observability, and the control/filter separation principle appears to avoid such issues as pole-zero cancellation, open-loop plant transfer-matrix inversion, minimum-phase open-loop plants, right-half plane transmission zeros, etc. In the ideal case of perfect model identification, the signal-processing poles are identical in the compensator filters considered as testable open-loop components and in the closed-loop controlled system. If maximally robust design procedures are followed, then in reality these poles shift minimally when the feedback loops are closed, provided that feedback of actuator signals is implemented separately from feedback of sensor signals by means of an active summing element, as demonstrated.< >