Fuzzy-type objective functons for NNDC

摘要

In this paper a "fuzzy-type" objective function is proposed for the training of Neural Network Direct Controllers (NNDC). The proposed method is based on cumulative evaluation of system's performance while the objectiv efunction seeks to minimize the "equivalent" system error, determined by fuzzy-set theory. The fuzzy-type objective function promises to increase convergence speed and perfomance of the NNDC. A successful application with respect to the robust control of a nonlinear dynamic system reveals a new approach for improved neural network capabilities extended by fuzzy logic theory.