Learning in movement and control

The authors propose novel knowledge representation and reasoning methods that are sufficient to develop a machine that can learn to control any controlled system in the same way that human beings learn: by observing only the input and output of the controlled systems. A simple implementation has been constructed to demonstrate the feasibility of building such a machine. It is not required that all the equations of controlled systems be known. It is only hypothesized that controlled systems can be described by a combination of several linear equations. The number of equations, the method of combination, and the parameters are acquired by learning. Simulation results are presented on the application of the proposed knowledge-based learning controller two-link and one-link systems