Optimal iterative learning control of an extrusion plant

Iterative learning control offers a method for the control of repetitive or cyclic processes which are encountered in the chemical, biotechnical, robot control and other areas. Successive control inputs are generated which asymptotically yield the desired run of the output. Whereas most of the work hitherto was concentrated on obtaining conditions necessary for convergence of control error, in the paper a method is derived in which the sequence of input functions iteratively minimizes a functional chosen to represent system performance. The derivations are based on the concept of the gradient of a functional which is expressed in terms of the Frechet differential. The application of the method for a control problem is illustrated with simulation examples and actual field tests.