Optimal learning control of mechanical manipulators in repetitive motions

Mechanical manipulators are often used to perform sequences of repetitive operations. A Learning Control method is presented in this paper that takes advantage of the repetitive nature of manipulator motions to improve trajectory tracking and speed. The manipulator dynamics are assumed to be completely unknown. The method requires only an initial but stable control point. The controller gains are modified in an optimal fashion until the manipulator performance is close to that of a reference model. The reference model is constructed in such a way as to track the reference input optimally. Control updates are performed off-line so that the manipulator operation and the Learning process have minimal interaction. This control method can be implemented in small computers since learning time is not very critical in repetitive systems. An overall systems architecture is presented in this paper along with simulations of a two-degree-of-freedom manipulator under learning.