Robust Model Free Fuzzy Adaptive Controller with fuzzy and crisp feedback error learning schemes

The aim of the paper is to investigate the different feedback error learning strategies used in conjunction with the Model Free Fuzzy Adaptive Controller (MFFAC). MFFAC guarantees tight control performance in the presence of disturbances and plant uncertainties. These uncertainties might arise due to the un-modelled dynamics of the plant under control or due to the different sensors introduced in the control loop at various points. The MFFAC is based on the model reference adaptive control (MRAC) philosophy and develops an inverse model of the plant by incorporating feedback error learning (FEL) strategy. The MFFAC is modelled as a fuzzy relational model and the identification scheme used is computationally undemanding. FEL governs the behaviour of the MFFAC during the learning phase i.e. whether the fuzzy controller will behave as a PI, PD or PID controller. It is an integral part of the overall identification scheme because it estimates the correct control signal which is consequently used to update the controller parameters. Fuzzy and crisp versions of the FEL have been studied and the comparison of the different approaches is discussed and their impact on the control performance is elaborated.