A nondeterminate controller with learning ability

A simple and versatile controller with the ability of learning from experiments is developed. The controller is nondeterminate because the number of attributes of the system and parameters of the control function are not pre-specified. Distinguished from previous learning controllers, this controller is more genetically behaved. Logic-based learning strategy is applied to the system. An optimal controller organization and its parameter tuning are obtained. The controller is particularly effective for control systems where the environment exhibits strong uncertainties and an exact mathematical model of the system dynamics is too complex to be managed by conventional means. Simulation results show that the controller is reliable over a wide range of operating conditions