Genetic evolution: a dynamic fuzzy approach

In this paper, a dynamic fuzzy approach is proposed for the selection of an evaluation function for a genetic algorithm (GA). The GA is in turn used to optimise a neural network (NN) architecture. A correctly classified pattern in the presence of error is considered for the fitness function. Fuzzy logic is used to dynamically select the chromosome for evolution. It gives a direction of evolution as well as provides more exploration among most desirable ones in the population, by dynamically changing the range of the membership function. To increase the resolution, different heights of the membership function with increasing heights towards more feasible features are considered. Modelling of a flexible manipulator is used to show the performance of the proposed approach. Results show that dynamic fuzzy logic performs better than fuzzy logic with fixed range and height.