Controller design based on orthogonal functions

The author presents a method of designing a learning controller based on the representation of the input-output signals by a set of orthogonal functions. Two theorems concerning the learning strategies in the design of the controller are given. They extend the results of S. Arimoto's (1985) betterment processes in the sense that some less restrictive conditions for convergence are needed. Examples are given to illustrate the results. Potential applications to robot manipulator control are noted.<
>