8.5 Review/Preview

In this chapter we have done several things. First, we demonstrated the ability of a GA to simultaneously select membership functions and rules, and to select real-valued consequent for the rules. Second, we introduced a fitness function applicable to constrained optimization problems. Finally, we discussed the effect of different implication operators on the performance when GAs were used for optimization.

At this juncture we have used a GA to optimize both the membership functions and the rules for three different control systems: (1) a liquid level controller, (2) a cart-pole controller, and (3) a satellite rendezvous controller. It seemed that our development was complete. However, it was at this juncture in the research project that a whole new world opened up to us. In the next chapter we will open a door to this world by repeating the comment our boss made to us one morning: “that cartpole controller works well, but what happens when the mass of the cart-changes?” In the next chapter we will introduce the idea of and need for adaptive fuzzy controllers.

References

Goldberg, D. E. (1989). Genetic algorithms in search, optimization, and machine learning. Reading, MA: Addison-Wesley.

Karr, C. L., Fleming, J. W., & Vann, P. A. (1994). Aspects of genetic algorithm-designed fuzzy logic controllers (Report of Investigations number 9515). Washington, DC: U.S. Department of the Interior, Bureau of Mines.