Adaptive neural network-based visual servoing control for manipulator with unknown output nonlinearities

In this paper, the problem of neural network control for visual servoing robotic system is addressed, where the unmodeled dynamics and output nonlinearity are taken into account simultaneously. An adaptive neural network module is constructed to approach the unknown dynamics, upon which, the robot dynamics are not required to be linearly decomposable and structurally known. The major superiority of this module lies in its conciseness and the computational-reduction operation. Moreover, the output nonlinearity is considered, and its undesirable effect is subsequently tackled without a prior knowledge of the model parameters in output mechanism. It is proven by the Lyapunov method that the image-space tracking error is driven to an adjustable neighborhood of origin. Numerical simulations and experiments under various situations are used to validate the performance of the proposed adaptive neural network based scheme.