A novel self-learning control algorithm for human-machine systems is presented. The designed controller is based on a probabilistic extension of recurrent fuzzy systems, which allows the consideration of non-deterministic information in addition to deterministic control signals. The behavior of the controller is adapted by varying the conditional probabilities of state switching, wherefore the automation-like structure of a recurrent fuzzy system is exploited. The adaptation is done by statistically evaluating the results from an objective and a subjective point of view. The developed transient probabilistic recurrent fuzzy controller (TP-RFC) considers two control objectives of different time scales. First, the actual control of the mechatronical subsystem and second, the consideration (self-leaning) of disturbances and the user’s idiosyncrasy in a long term. An application of the proposed TP-RFC to a washing machine is shown by simulation.      «

A novel self-learning control algorithm for human-machine systems is presented. The designed controller is based on a probabilistic extension of recurrent fuzzy systems, which allows the consideration of non-deterministic information in addition to deterministic control signals. The behavior of the controller is adapted by varying the conditional probabilities of state switching, wherefore the automation-like structure of a recurrent fuzzy system is exploited. The adaptation is done by statistic...     »