Can a two-link manipulator learn how to write?

This paper considers a version of the problem of how to teach robots to write characters in an actual environment. In particular, a feedforward controller is designed for two-link manipulators to improve tracking performance despite limited knowledge of the surroundings. An adaptive scheme, called MIMO-FEL (Multi-input Multi-output Feedback Error Learning), is employed to achieve the objective. After convergence, the feedforward controllers are switched depending on the target character to be written. This switching strategy is a clear contrast with the precise identification approach, which uses a single general-purpose controller.