Model based tuning and adaption of fuzzy logic controllers

This paper presents a fuzzy logic based control structure enhanced with supervised learning and/or adaption functionalities. Availability of at least a partial process model is assumed. Nonlinear process identification procedure is used to complete the partial model. Based on this process identification model using the techniques of systems sensitivity theory, the necessary gradients are generated to guide the training process and thus to keep the training time (the number of observations) at minimum. The process identification and the controller tuning can run in parallel, in this way the online adaption of the controller can be realized in a straightforward way. A supervisory robot control problem is shown to demonstrate the capabilities of the scheme proposed