Parallel dynamic switching of reasoning methods in a fuzzy system

Different fuzzy reasoning methods show different performance depending on the reasoning situation. The best fuzzy reasoning method is decided not only by the application task but also by the time variant reasoning situation. A dynamic switching fuzzy system (DSFS) is proposed which chooses the best performing combination of available fuzzy intersection operators and defuzzification methods according to the reasoning situation at that point in time. DSFS uses 80 different reasoning methods connected in parallel and the best performing one is dynamically chosen using a criterion of target prediction strategy involving fuzzy constraints and fuzzy goals. Comparisons in balancing a simulated inverted pendulum using a modern controller, several conventional fuzzy logic controllers (FLCs), and a DSFS FLC were carried out under different conditions such as normal, degraded rules, errors in scaling, windy conditions, etc. DSFS was found to have higher performance, better robustness, and better fault tolerance