Closed-loop identification of restricted complexity models for feedback/feedforward control using iterative refinement

A novel technique for closed-loop identification of reduced-order models for combined feedback/feedforward control is presented. For this purpose identification and control design are treated as joint problems for finding models whose intended use is the design of high performance compensators. The problem is solved by an iterative scheme of repeated identification and control design. The scheme involves the following steps: 1) injecting a dither signal at the manipulated variable and collecting closed-loop (u,y) data, 2) prefiltering the closed-loop data to emphasize the frequencies most important for control and 3) low order parameter estimation and control design.