Flexible-joint manipulator adaptive control based on recurrent Elman neural networks and dynamic surface approach

For overcoming the nonlinearity, uncertainty and unknown external disturbance in the model of flexible-joint manipulator driven by surface-mounted permanent magnet synchronous motors(PMSM), an adaptive dynamic surface control(DSC) approach is proposed to design a position-tracking control system in the joint space. Control laws are obtained from DSC technique, which reduces the complexity of backstepping controller. The uncertainties of the model are observed and compensated online by the recurrent Elman neural networks(RENNs). And the adaptation laws of RENNs' weights are induced from the Lyapunov stability analysis. The simulation studies show that the proposed method provides a good robustness against payload uncertainties and external disturbances, and the position tracking performance is improved greatly comparing with the conventional DSC method.