Design of a neural-fuzzy controller based on fuzzy differentialcompetitive learning

摘要

In this paper, a novel neural-fuzzy controller based on fuzzydifferential competitive learning is proposed. Since one of the mostimportant parts is the generation of the fuzzy rules in the design ofthe fuzzy control system, a fast learning algorithm, fuzzy differentialcompetitive learning (FDCL), for the generation of the rules is appliedin the fuzzy control system. The FDCL algorithm adopts a principle oflearn according to how well it wins. Unlike the previous competitivelearning algorithm such as crisp competitive learning algorithms whereonly one neuron will win and learn at each competition step every neuronin the neural network based on FDCL algorithm will along with itsdifferent distance to the input pattern and learns the patternaccordingly. Compared with the ordinary competitive learning algorithmthe proposed FDCL algorithm has the various distinguishing features. TheFDCL algorithm is implanted in the neural network based fuzzy system andthe network adopted is fuzzy associative memory system (FAMS) whichsimulates the knowledge representation and inference process by usingfuzzy notation and by association in neural networks. In FAMS the fuzzyrules will be generated by clustering the input-output training datathrough the FDCL paradigm. By using the FDCL algorithm the neuralnetwork can highly refine knowledge and represent the expert experience