Fuzzy predictive control based on human reasoning
Robert Babuška, João M. C. Sousa, H.B. Verbruggen · Data Archiving and Networked Services (DANS) · 1995
Human knowledge is an important source of information for modeling and control of complex dynamic processes.Fuzzy sets proved to be suitable for dealing with subjective uncertainty encountered when incorporating human knowledge in the design of automatic control Systems.Besides direct fuzzy control, in which the control law is explicitly described by If-Then rules, the knowledgebased approach can be applied at a higher level for formulating the control objectives and constraints.Appropriate control actions are then found by means of a multistage fuzzy décision making algorithm, using optimization over a finite horizon as in conventional prédictive control.Compared to the standard quadratic objective fonction, the knowledge-based approach gives the designer more freedom in specifying the desired process behavior.By using fuzzy models, the uncertainty arising from the modeling of complex and partially unknown Systems can be represented at the same conceptual level as is the uncertainty in the goals and constraints.Finally, a model-based search for an optimal control strategy can be combined with model-freereinforcement techniques inspired by human learning.