Adapting Gain and Sensibility of FLCs with Genetic Algorithms

Fuzzy Logic Controllers are knowledge-based systems, incorporating human knowledge into their Knowledge Base through Fuzzy Rules and Fuzzy Membership Functions (among other information elements). The deenition of these Fuzzy Rules and Fuzzy Membership Functions is actually aaected by subjective decisions, having a great innuence over the performance of the Fuzzy Controller. This paper describes the application of genetic strategies, with a learning purpose, to the Knowledge Base of Fuzzy Controllers. The evolution is applied to modify the gain and sensibility of the controller (through the scaling function of each input and output variable), and the Rule Base. Fuzzy Logic Controllers (FLCs) are knowledge-based system characterized by using knowledge represented as a set of rules and membership functions, and a reasoning strategy based on the aggregation operators, the fuzzy connectives and the inference method. The main concepts related to FLCs are described in 1, 2]. A possible modular structure of an FLC (Figure 1) is: 1. An Input Scaling module that maps input variables to a normalized universe of discourse for each input. 2. A Fuzziication interface that converts the values of normalized input variables to fuzzy information. 3. A Knowledge Base. Figure 1: Structure of an FLC. 4. An Inference Engine that infers fuzzy control actions employing fuzzy implications and the rules of inference of fuzzy logic. 5. A Defuzziication interface that yields a non fuzzy normalized control action from an inferred fuzzy control action. 6. An Output Scaling module that maps the normalized outputs to nal outputs. As in any other kind of knowledge-based control system , one of the most important questions when deen-ing or implementing a Fuzzy Logic Controller to solve a certain control problem, is the derivation of the Knowledge Base. The performance of the system is strongly related to the quality of the Knowledge Base, and consequently with the process of knowledge acquisition. In some cases, this process does not take place because there are no experts available in the eld of a certain problem. It is possible too that the acquired knowledge only represents an incomplete or a partially incorrect description of the control algorithm. When the available knowledge is not considered good enough to be used as the control algorithm of the system, some kind of learning or adaptation is needed. With this aim, ideas arising out of two main areas have been applied: ideas coming from Artiicial Neural Networks (ANNs) 3] and …