How to design an adaptive fuzzy decision-making model?

The design process for fuzzy controllers that is created based on the utilization of experimental data from human experts has been successful in many applications. Moreover, the approach to building fuzzy controllers via numerical input-output information is progressively finding use. Despite which method is implanted, several challenges are faced with practical control problems, including the following:

• The design of fuzzy controllers is operated in an unexpected way so that it is often challenging to select at least some of the controller model parameters. For instance, it is usually difficult to recognize how to choose membership functions and rule-base to achieve a particular required level of performance.

• The fuzzy controller formed for the nominal plant might later perform ineffectively if considerable and unpredictable plant parameter deviations happen, or if there is noise or some other kind of disturbance or another type of environmental consequences. Thereby, it may be complicated to perform the preliminary synthesis of the fuzzy controller, and if the plant varies as the closed-loop system is performing, we might not be able to preserve acceptable performance levels.

On the other hand, a learning system demonstrates the capability to stage its performance over time by interacting with its surroundings. A learning control system is proposed so that its learning controller has the capacity to improve the performance of the closed-loop system by creating command inputs to the plant and exploiting feedback information directly from the plant. In this method, the adaptation mechanism detects the signals from the control system and adjusts the parameters of the controller to support performance even if there are variations in the plant.

Occasionally, the target performance is described with a reference model, and the controller then tries to make the closed-loop system work as though the reference model would even if the plant alters. This is called model reference adaptive control (MRAC), as demonstrated in figure below.