Neuro-fuzzy architecture for identification and tracking control of a robot manipulator

Neural networks and fuzzy systems have been applied very successfully ill the identification and control of dynamic systems. This paper presents combination of the fuzzy logic controller and neural network identification structure, integrated into robotic system, to provide extensive capabilities. We first discuss the fuzzy logic controller (FLC), describe its main components such as fuzzifier, fuzzy rule base, fuzzy inference engine and defuzzifier. We then focus Oil the neural network (NN) plant model, trained on-line using the backpropagation training optimization algorithm with an adaptive learning rate. The optimization algorithm is performed at each sample time to compute the optimal control input. The results confirm the effectiveness of the proposed identification and control architectures.