Evolutionary fuzzy speed regulation for a DC motor

Fuzzy logic is a powerful tool in control of systems with ill-defined, inaccurate or unknown mathematical models. In classical applications of fuzzy logic, however, there is a great dependency on proper expert knowledge acquisition. The authors remove that dependency by using a genetic algorithm (GA) to automatically determine parameters of fuzzy rule sets such as membership functions. This approach differs from conventional applications of GA-fuzzy knowledge development in that expert knowledge is incorporated in creating an initial highly fit population while allowing for randomness among members of the population for diversity. This method is useful for search in GA-hard landscapes and is successfully applied to speed regulation of a DC motor. It is shown that the presented method improves upon the initial fuzzy knowledge-base and significantly outperforms classical PID response.