An adaptive fuzzy control system for robotic manipulators

In this paper, an adaptive fuzzy control system is developed for robotic manipulators whose mathematical models are unknown. This control system is constructed by combining three methods, including an independent joint control strategy, to generate initial rules by using both data and linguistic information, and online parameter optimization through learning. The independent joint control strategy reduces the number of fuzzy rules, sample data and human experience provides information to construct initial rules which serve as initial values for parameter optimization. Parameter optimization is a key step exerting the approximation ability of adaptive fuzzy systems to compensate for the disturbances caused by the coupling effects and revising the incorrect components of the experiences used to construct initial rules. The backpropagation algorithm is applied due to the structural equivalence between adaptive fuzzy systems and multilayer feedforward neural networks. Initial rules help to improve the convergence ability of BP algorithm.